Mar 8, 2026 Enterprise Brain Arch and Next Steps - Transcript 00:00:00 Gopal Gottumukkala: That is cool. Vara Kumar Jagarapu: Excuse Gopal Gottumukkala: All Vara Kumar Jagarapu: me. Yeshwanth Reddy Yerraguntla: Hello. Hi. Gopal Gottumukkala: right, Yeshwanth Reddy Yerraguntla: Hi. Vara Kumar Jagarapu: Andy, are you Yeshwanth Reddy Yerraguntla: So, goals are free. Vara Kumar Jagarapu: open? Yeshwanth Reddy Yerraguntla: Will you be able to spend I'm sharing my Gopal Gottumukkala: there you go. Yeshwanth Reddy Yerraguntla: screen? So, Gopal Gottumukkala: Sure. Yeshwanth Reddy Yerraguntla: the agenda is uh two things. one is uh to walk you through the architecture itself. Second thing is uh how we have broken down the tasks and uh uh like we are sort of uh not clear on how to break down the tasks Gopal Gottumukkala: I didn't have an option. Yeshwanth Reddy Yerraguntla: uh so as to give to the juniors because you are dealing with a team of 10. how to properly separate out all the tasks so that everyone can work in parallel with minimum dependencies on each other. Gopal Gottumukkala: Got it. Yeshwanth Reddy Yerraguntla: Yeah. I you can always interrupt me when you think you u are clear on something but I'll anyway start from scratch because um like I don't know how much uh uh was like changed from the last time we discussed so first thing is uh it's already it's already getting record. 00:05:03 Gopal Gottumukkala: Okay. Vara Kumar Jagarapu: Okay, thank Yeshwanth Reddy Yerraguntla: Yeah, sorry. First thing is there are um mainly four individual agents uh that are in picture. Okay, even before that the whole point of this architecture is this has to be uh what is it proactive more than reactive and uh um for proactivity we had to introduce a few additional features some things are straightforward some are new in even AI terminology and AI like there's nothing standard in certain aspects so it's going to be something we'll invent or discover by ourselves. Anyway, so there are four main sources of intelligence in the whole system. The wherever you see this icon, right? This blue icon, I'm saying that is uh that is a place where a lot of uh AI uh is necessary. So first uh overall like larger boxes user interfaces where um um user has uh three different sources of interaction, one is the dashboard, one is the chat interface and uh one is not a source of interaction but uh whenever there are critical alerts they the user will directly get them via email and uh whatever phone um dashboard and the chat interface um chat interface I think uh you already know how it looks like but if I have to take a small example um yeah 00:07:04 Yeshwanth Reddy Yerraguntla: so the way we have currently built is um an interface where it's not only giving us text but it is also giving us flowcharts in the form of mermaid diagrams. It can give the standard mermaid u artifacts like tables uh headers bold all those things but it also can give uh u uh plots uh dynamically and the technology we are currently using is plotly pl which is basically uh an interface where python the agent will write python code that code is sent to plotly's uh uh library that will generate an iframe and it will that iframe we are directly showing here. So whatever plotly supports they are naturally allowed showed here. Uh next thing is uh yeah mainly these are this is how the chat we are looking at. Um coming to the dashboard itself it should look something like this. This is a very like an alpha of an alpha kind of a thing, right? So we will have widgets on one side, we'll have chat on one side. 00:08:26 Yeshwanth Reddy Yerraguntla: Ideally speaking, the widgets will uh show what is important for me to consume on that day. This is like my newspaper. Uh and uh whatever is happening, whatever is latest in Diwami, according to Mails, according to Jira, according to Google Drive, everything should show up here automatically. Uh I can I should always be able to select whatever. Um so I'm saying link approvals to whatever um so it was able to connect a page with this guy. So that kind of crowd operation should be possible on the dashboard itself. Uh right chat should able to create uh read update delete these widgets and inside widgets you'll have more widgets like you are maintaining widgets of widgets of widgets kind of an hierarchy. It's just a concept. We still have not get gotten clarity on is this what we need or if we need this how to put the designs but we are reasonably sure that at least 50% of the concepts will come into picture. So that is dashboard. 00:09:49 Yeshwanth Reddy Yerraguntla: We just discussed chat interface. We have alerts and notifications. Now alerts apart from this one guy the dashboard and chat interface need to be updated directly via visualization intelligence and uh the way it should work is uh someone send it what is the content um and the the responsibility of this agent is mainly to uh respons responsibility is to decide on how to visualize it, right? Should I use a chart? Should I use a plain text? Gopal Gottumukkala: Thank Yeshwanth Reddy Yerraguntla: Right? There are a lot of decisions this itself needs to take uh based on good Gopal Gottumukkala: you. Yeshwanth Reddy Yerraguntla: practices and based on user preferences also. Right? So that itself is a lot of burden on this uh intelligence. So the content is going to be received by this agent from somewhere else. it will only decide what is the best way to show it on the user interface. So that is one agent. So going a little downstream we have core agent and user aware agent. 00:11:06 Yeshwanth Reddy Yerraguntla: So this red box basically is this guy where you have two different sources of intelligence. At a very high level again um user aware intelligent is just consuming raw news uh and it is deciding on how to distribute the news to every person in the organization because the same uh uh you know update about uh Tata Steel everybody needs to know a different aspect of it. What Pratiman madam needs to know is different from what Naven needs to know different from what the tech lead needs to know. Same news. So this agent is deciding on that user awareness and deciding for each user what news to publish. Whereas real time agent is deciding what is the content. Gopal Gottumukkala: There we go. Yeshwanth Reddy Yerraguntla: So this is doing the heavy lifting of uh getting all information from every data source. It is getting uh its own uh existing uh information. Right? So currently or maybe yesterday what was the news for Tata Steel uh it is it is synthesizing that yesterday's news with today's updates. 00:12:24 Yeshwanth Reddy Yerraguntla: So it is constantly updating its own source of intellig knowledge while it is publishing from yesterday to today. These are all the things that happened or last five days. This is how uh every aspect of Tata Steel has changed right. So this is a very detailed sort of like a sort of like a journalist uh role right. It is just going to the ground fetching what is happening where and uh it is just sending it to the editor where editor is deciding uh uh for every person in the organization how to write the news and this guy is deciding for every person based on their preferences and best practices how to publish it in a way that they consume it uh in the most uh scientific and easy to digest manner. So this is the uh main pipeline data agents are the whatever they they do their job of every day what has been happening I it will constantly update real time intelligence agent uh to some extent it is they are also doing other things like uh rback it they are doing stuff like um maintaining uh some kind of a c so that if a user is asking something u something simple it it it if it is in case it will directly fetch the answer some kind like few basic functionalities this these agents are taking care of um right so this is sort of the pipeline but this pipeline itself can be split into two different 00:14:14 Yeshwanth Reddy Yerraguntla: uh aspects one aspect is when a user is asking a question like for example in this case right I ask a question on the user interface that question directly goes to real time uh decision intelligence agent what I mean is this chat interface is directly sending a query to ad hoc queries in the form of ad hoc queries to um real time intelligence this is deciding uh of of the 20 agents it is connected to whom to route Rotate to or which five agents to route to. Right? It's a uh one or all kind of a situation. So this is deciding for every agent what question to ask then get back the answers and then send it back to UI. Uh so that is one one uh side of the whole highway uh that we are we are trying to build. The other side of the highway is there has to be a constant refresh so that my dashboard is constantly up to date. So what that there what happens is uh refresh agent is the heartbeat in this whole mechanism. 00:15:38 Yeshwanth Reddy Yerraguntla: So this will send a ping to realtime agent saying uh from all these agents you fetch the news. So naturally it will do the same fetch give me all the news articles from yesterday or last 6 hours. It will get back the information it will update its knowledge base. It will send the news to user aware agent. This will that pipeline will continue again. So all I'm trying to say is there are two sources of triggers. One trigger is directly from the user uh which is an ad hoc basis. One trigger is from the agent refresh mechanism which is where the proactivity comes into picture. the fact that when I wake up tomorrow I see a fresh dashboard the the way it is h happening is because this guy triggered in the midnight what is the so I'm just giving examples um so far u is it clear goal sir Gopal Gottumukkala: Yeah, I'm good as so far. It is clear not deviated anything from earlier what we 00:16:53 Yeshwanth Reddy Yerraguntla: Yeah. Right. Gopal Gottumukkala: discussed. Yeshwanth Reddy Yerraguntla: Um so then the uh question is without going too much into details. Um I will give you like a very high uh level of what are the like somewhat high level responsibilities. One thing that is also a very important feature in our uh enterprise brain is training lessons. Best practices uh need to be taught to the whole system so that every organization can teach the whole system to adapt it to the organization. So standard operating procedures is one um example when they change on a quartertoquarter basis someone can simply code the UI and change it right so that's kind of a training lesson mechanism we have to bring into picture training lessons is a function of sorry not a function training lessons can happen both at the organization level and at the data source level where are training lessons right so it depends whether something being taught is agent specific or uh organization specific. One more important feature that uh uh is very much in line with the same training lesson is user preferences. 00:18:25 Yeshwanth Reddy Yerraguntla: Sometimes users will say from tomorrow or next time uh I prefer bar charts over something. So that needs to go sit in some other database. One more uh uh very key crucial uh data basis how should we manage the organizational uh news I'm calling it snapshot so on a day-to-day basis or hour to hour basis how uh the the processes are evolving at a user level at a project level uh at a pipeline level whatever is the entity for any organization you'll have those constant flow of things, right? So for us uh in software engineering they can be projects, there can be hiring uh uh pipelines, they can be leads to some extent leads are also part of projects but for some other company this might change. It might be a supply chain entity could be whatever. So uh point is somehow we need to manage a very high refresh uh uh database that is the main purpose of this is to avoid answering questions uh every time I don't need to go to data agents for every little thing right so if if navinsar is coming and asking what's going on in data steel uh it is such a common question or such an important question that we'd rather store everything about Tatast steel in one place so that uh I'm 00:20:06 Gopal Gottumukkala: Okay. Yeshwanth Reddy Yerraguntla: saving a lot of round trips right so that uh yeah that is one more data source of information source of information that uh is crucial is the usual LLM M knowledge itself like what is the LLM's purpose? What is it even doing in the whole system? Instructions. Um it we are calling some things as runbook some things as skills. Skills are nothing but Python functions so that it can call them directly. Uh guard rails resources resources could be static material like notion documents right. So things which don't have a large I mean these are these are things that don't refresh usually at a high level. So that is also one source of information both for the core agent as well as the data agent. Um so this is how we are trying to organize the whole uh we just discussed the pipeline of information travel and we discussed how is information being stored in rest. So what we have done goals is we uh I mean for every one of these big boxes we went through and understood what are the life cycles associated with those boxes right so this is how we started what are the components that the these life cycles involve what are the data packets traveling and what are the features expected from those uh life cycles. 00:21:54 Yeshwanth Reddy Yerraguntla: So this is how we were able to come up with this table of what we want to build right. Um so till here I think we are clear on what we need to build. Um to some extent we also went uh uh did some very basic low-level design of how common component should look like because there are four agents um four sources of intelligence. We thought every uh every agent will have this common functionalities, right? So it's still not clear how exactly they will be. But we thought there has to be some initialization file. There has to be some location where it can fetch all its tools from uh some way to initialize uh a posgress database and that database should contain all these uh tables and columns. basic commonalities, right? So, all we uh were able to understand from our diagrams was okay, there are these few number of agents where every time an agent is going to be initialized uh while initializing these are the things that it needs and 00:23:10 Gopal Gottumukkala: Okay. Yeshwanth Reddy Yerraguntla: uh database should be in this format kind of a thing. So beyond that we did not go too deep into low level because um uh like I don't know if we go too deep I thought we might waste a lot of time uh understanding the nitty-g gritties and uh we wanted somewhere to figure out if we can uh hand over this from here to the juniors and uh uh seniors also. Um right so so far Wah me Rajakha and Manisha mainly were in the process of generating uh this much information we went into trying to come up with a plan but u maybe it's just me that uh I'm not able to like completely put it uh properly on paper maybe I'm missing where if there is any better way of representing it but we are sort Um stuck here that is what it is. Go pulser V. Do you want to add anything? V or Manisha. Vara Kumar Jagarapu: nothing. So let's go further. So and a few things and whatever the life cycles we have what are the dependencies and all the things with some acceptance criteria we listed down in this sheet and uh yeah every life cycle then we thought of going into more details just a developer point of view if they can discuss that approach and if we is there any blockers and all those things like we creating your tasks. 00:25:14 Vara Kumar Jagarapu: So this is how we thought of going and maybe if we can get some more betterments on top of it for the deliverables. Gopal Gottumukkala: I a thought process on it but uh see this is natural that we'll get stuck in this stage in any project especially the problems which were not solved earlier either by us or by somebody else we are trying to create a solution this is Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: natural this is where we'll probably reach here and there and then you know next steps will be a little punchy so what I was thinking last last week when I met you also if we are trying to go deep and then trying to get tasks laid out for a long period it will be difficult the these this I mean this level of detailing is good enough we have to take a pause Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: here and put a thin line of end to end uh flow ignoring that how many agents are going to come basically my thought process is there is a front end there is a middleware there is the 00:26:26 Yeshwanth Reddy Yerraguntla: H Gopal Gottumukkala: back end just like a regular MBC what fix right Yeshwanth Reddy Yerraguntla: Right. Gopal Gottumukkala: put shells don't try to bulge the shell yet say if I have to elaborate take one data source or two data sources which are already done Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: then you take already done data sources also but converting the data source putting a agent Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: wrapper don't do it and all just take one so what my goal Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: is I have to understand let me take drive itself right and I have to I have to acquire the knowledge out of the drive and keep it ready and then another agent is running and saying that okay something needs to be verified for somebody so maybe it is going to tell me that a new meeting recording came and out of that there are some action items came I'm just hypothetically creating a case right wherein architecture has to be changed. That is one of the action items. Somehow it came right. Then either I should get a notification or when I landed there my dashboard should say say that you know you have an open item where you need to respond. 00:27:41 Gopal Gottumukkala: If I go there and ask okay what exactly could you elaborate then probably I should be pointing Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: to the transcript of that meeting. Okay. And I could ask further. Okay, what exactly people are saying? I mean, which piece needs to be changed or could you get me the relevant information? So, it could probably point me to the old architecture that needs to be changed and the spec documents things like that, right? So, basically I am able to traverse through the information. I got the first hit and then I'm going deep. Right? That completes the entire cycle. Yeshwanth Reddy Yerraguntla: Okay, you are saying keep it lean but try to fix the whole cycle at least for one or two use Gopal Gottumukkala: Yes. Yeshwanth Reddy Yerraguntla: cases, Gopal Gottumukkala: So what happens is usually in this start process your DBs will be stabilized. Yeshwanth Reddy Yerraguntla: right? Gopal Gottumukkala: The only clumsiness it will be is the borders between the agents will be little still clumsy because we have not explored all the possibilities of agents, all the possibilities of data sources. 00:28:55 Yeshwanth Reddy Yerraguntla: Got Gopal Gottumukkala: We don't know where to draw what, right? Which is okay. As long as we we know in mind that you know these borders are a little thicker. Yeshwanth Reddy Yerraguntla: it. Gopal Gottumukkala: We have to make it thin over a period of time. As long as we keep that in mind, one cycle gets us the complete picture of what are the possibilities, Yeshwanth Reddy Yerraguntla: All Gopal Gottumukkala: what are not the possibilities. We might not be able to solve all the cases in this one line because there are dependencies across agents, right? That's the whole point here. Yeshwanth Reddy Yerraguntla: right. Gopal Gottumukkala: But few preferences don't worry about too many. Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: Uh we can put simple rules like you know I always wanted to get sequence diagram rather than a flowchart. That's my choice right wherever it is at sometimes you know only a flowchart works Yeshwanth Reddy Yerraguntla: All right. All Gopal Gottumukkala: then flowchart is fine but as long as there is a possibility of uh writing a sequence diagram my choice is sequence diagram and I always concentrate on technical task even if there is a project management task coming on to me I usually 00:29:53 Yeshwanth Reddy Yerraguntla: right. Gopal Gottumukkala: do who could be the uh right fit to hand over the task so assume that create one personality like that and we will try to target that personality. Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: Everything else what we do in other agents is if this agent requires something we'll put in another agent hard code don't solve that agent we know what it is going to give right so we'll put that kind of response from there so that way we'll concentrate into Yeshwanth Reddy Yerraguntla: Yeah. Okay. Gopal Gottumukkala: one flow of what is happening in the system that gives us the entire picture of no now we look into each touch point and then see what is our bigger dream and how deep that particular touch point is. Now we know the clearcut task that we need to put. Yeshwanth Reddy Yerraguntla: Okay, I Gopal Gottumukkala: So what we are going to do on the first one is a is a kind of very loosely Yeshwanth Reddy Yerraguntla: see. Gopal Gottumukkala: connected tasks but only connectivity that we are trying to establish is an end to end 00:30:56 Yeshwanth Reddy Yerraguntla: I see. Gopal Gottumukkala: flow somewhere we are reading the information acquiring the knowledge some Yeshwanth Reddy Yerraguntla: I understood. Gopal Gottumukkala: kind of you know uh to alert is coming in then that is being Yeshwanth Reddy Yerraguntla: Got it. Gopal Gottumukkala: presented on the UI and we are allowing the user to talk on top of it reach to Yeshwanth Reddy Yerraguntla: All right. Got Gopal Gottumukkala: So their ends are cycle. Yeshwanth Reddy Yerraguntla: it. Gopal Gottumukkala: Actually everything else is bulging all the you know uh wherever we is joining something I call them as nodes or you know touch points that touch point needs to be elaborated. Yeshwanth Reddy Yerraguntla: Right. Gopal Gottumukkala: Now we know which touch points seriously needs to be elaborated for the first B first business goals Yeshwanth Reddy Yerraguntla: Got it. Gopal Gottumukkala: right somewhere chart interface suddenly becomes so serious then we'll concentrate in that Yeshwanth Reddy Yerraguntla: Right. Gopal Gottumukkala: and then try to uh increase the depth on it again same approach there also we don't try to solve the 100% problem of chart interface we'll try to solve the problem of chart interface in the same road we don't change the road map yet the key is we are not changing the road map what we are trying to attain is not the entire depth but what is currently being regret without hampering the thought process. 00:32:13 Yeshwanth Reddy Yerraguntla: Makes sense. Gopal Gottumukkala: So that way what what will be next one month looking like is every week we'll have tasks but as of this week task only 60 maybe 50 60% tasks are solid everything else is like very hazy which will get by the end of the week or for probably for the next week and then again cycle repeats cycle repeats what changes is this week we might be concentrating on one thing next week we might be concentrating on next Yeshwanth Reddy Yerraguntla: Okay, I see what you mean. Gopal Gottumukkala: that that way we'll get clarity otherwise what happens is we are trying to solve the problem 2D then we are not proving the point that this problem is solved Yeshwanth Reddy Yerraguntla: Yeah, that is true. A this it sort of became a chicken and egg problem and we were struggling. Now I have understood what you're trying to Gopal Gottumukkala: let's pick one yeah let's pick only on everything else Yeshwanth Reddy Yerraguntla: say. Gopal Gottumukkala: we'll stop it out. I am not saying don't do it. 00:33:17 Gopal Gottumukkala: What I'm saying is if we need four agents, Yeshwanth Reddy Yerraguntla: True. Gopal Gottumukkala: we'll put four agents. Only one agent we actually could all other agents are going to return the demi responses the way we Yeshwanth Reddy Yerraguntla: Cut Gopal Gottumukkala: want. Yeshwanth Reddy Yerraguntla: it. Could it? Yeah. Yeah. Gopal Gottumukkala: Then it gives two different uh necessities clarified. One is from a solution and core point of view how the structure should be and it also gives us a better idea and how do we need to orchestrate in the infra also Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: I I'm I'm thinking one flow will get us a at least you know 40 50% of the architectural footprint right from solution core and infra Then we know what pieces to concentrate and what pieces to bulge. Yeshwanth Reddy Yerraguntla: Yeah, I think I'm able to follow what you're saying. If I were to like concretize what you're saying, right? Let's not try to think of this as the whole uh list of things that we have to do. 00:34:37 Yeshwanth Reddy Yerraguntla: Um I don't know like you can correct me if I'm going in a different Gopal Gottumukkala: Yeah. Yeah. Please go Yeshwanth Reddy Yerraguntla: direction but what I'm thinking is like you said try Gopal Gottumukkala: ahead. Yeshwanth Reddy Yerraguntla: to have a uh where is this in few of the life cycles are very very important at least to begin with like one life cycle very important is ad hoc questioning and answering where I am asking a question uh from the UI it is flowing to these agents These agents are asking data source agents. Data source are giving back answers. Now in this entire life cycle it is there are like four or five touch points which we don't have clarity on anywhere. So once we take up this activity we will be in a point where after finishing this at least 20% of the whole architecture we will have clarity on and because we are attemp we are attempting this only because it has high Gopal Gottumukkala: Yeah. Yeshwanth Reddy Yerraguntla: business value like in the initial days 00:35:42 Gopal Gottumukkala: Okay. Yeshwanth Reddy Yerraguntla: then once that sort of like a backbone is ready around that backbone you can start picking other life cycles that are also important and easy to uh easy to implement given the backbone and slowly um uh nourish the whole uh uh Gopal Gottumukkala: Yeah. See if you look at this right all this can be uh you know Yeshwanth Reddy Yerraguntla: code. Gopal Gottumukkala: prioritized on the uncertainity of implementations right see alerting life cycle I'm not worried what is more important is some agent Yeshwanth Reddy Yerraguntla: Yes, it's Gopal Gottumukkala: running behind the scenes and it is trying to figure out what these guys are looking for based on that it look look into the knowledge and say okay because these guys are looking into these five perspectives let me pull out anything that needs to be alerted. Yeshwanth Reddy Yerraguntla: H right. Gopal Gottumukkala: That is one cycle right in those there could be all across the Yeshwanth Reddy Yerraguntla: Yeah, cut All Gopal Gottumukkala: company there could be 30 40 concentrated different areas where people might be looking at Sacha looks at differently Praakaru looks differently we only pick one or two say probably we'll take Navin's perspective or Brahma's perspective all other perspectives we'll keep it aside it proves a point that okay something is always looking into the knowledge and then throwing the alerts In my mind that problem is solved. 00:37:07 Yeshwanth Reddy Yerraguntla: right. Gopal Gottumukkala: Now another big problem is training pick the same use cases Yeshwanth Reddy Yerraguntla: H Gopal Gottumukkala: wherein what Naven is looking for what system is talking about is every time Naven is asking what is the status he is it is simply saying you know you burn uh 1k out of 200k but Naven's intention is how much of the task got burnt up. So now we can teach okay so by default don't look into the you know burn rate by the money rather than you look into number of features not so ideally in next time it should be looking that okay you are supposed to uh close down 20 out of 200 tasks but you only knock down 15 out of uh 200. earlier perspective will also come but it'll come a little later in his list of things to know. So that proves a point of training because training has any Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: number of variance right it could be how to calculate the revenue or something like you Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: know how to calculate a particular person's burn rate or you know velocity we pick one case whichever wherever we have clearcut data a target should be can be proved out 00:38:32 Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: of the priorities that we are thinking where and all we have the data solid database right see if we would have already integrated a Jira by then ideally anything related to tasks how many tasks being closed per day per week or towards a milestone would be easy questions to answer and we can easily prove also whether the agent is uh demonstrating the right answers or not right so we all have to keep I mean if we have to follow this approach. We all have to keep in mind that we are targeting solidity in the sense whatever the piece that we are trying to do we should be able to feel that is right or wrong because inherently what we are trying to prove to ourself is are we in the right path is our uh connected lines are being established or not. one, Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: two or three lines. If we can establish, we all have a pattern of how to conquer the remaining life cycles. Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: We we might go here and there, we might deviate among ourselves in the first lane itself. 00:39:49 Gopal Gottumukkala: That is also natural. But some decisions we would have made and corrected our path, right? We all have to participate in those decisioning so that the next cycle already all of us know what to think or how to think. Then by the third one you can paralyze because now you have more than one or two people who can actually take the lanes till the end without Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: needing every one minute coordination. Anyway, this is not my theory. This is how we actually try to crack a problem when we when we got an idea feels like it is right. But uh we have to prove the point. So we can pick one or two H1 and just try to put an end to end lines whichever feels more practical to implement we'll pick Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: that one out of those two. Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: So at a very high level I am not worried about number of data sources. That problem we already solved many times right not together but independently. 00:41:14 Yeshwanth Reddy Yerraguntla: Cut it. Cut it. Gopal Gottumukkala: So putting things together is not a problem because you already uh created cross Yeshwanth Reddy Yerraguntla: product. Gopal Gottumukkala: communication between LLMs that is also not a problem but Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: training methodology was proven in a very structured manner in uh optim Yeshwanth Reddy Yerraguntla: Part two. Gopal Gottumukkala: right here it is very vague it is not that solid because there the use case is Yeshwanth Reddy Yerraguntla: Yes. Gopal Gottumukkala: very constrained training will become easy you know what to teach Yeshwanth Reddy Yerraguntla: Yes. Gopal Gottumukkala: here it is a little hazy Okay. Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: So probably training could be the next best best best bet for us. I'm saying that the training will be conquered and solved in a week. But at least it gives us an idea of how to think many number of ways of training that we need to train. So we know what to do. uh the architecture that you put in right in my mind that Yeshwanth Reddy Yerraguntla: Mhm. Gopal Gottumukkala: orchestration talking to different agents has many many layers on top of each other in every aspect say training right training will have n 00:42:22 Yeshwanth Reddy Yerraguntla: Hey Gopal Gottumukkala: types of trainings again so a central training engine has to coordinate with different types of trainings and push it to the appropriate agents in my mind that's how I'm thinking even alerting is also similar philos You will have a centralized alerting intelligence. Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: Then there will be concentrated alerts uh alerting capable agents or engines whoever we want to call they only concentrate on certain area. The central agent is going to know that you know how to read them because multiple data points are going to say the same thing right because I got a mail that mail is driven uh mail has a meeting link and also a document all the three are talking about the same thing there is a problem you you said it is event driven but customer decided event Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: driven they'll want to go out after 6 months so architecture has to be changed all the three are basing the same kind of alert Somebody has to sit on top of it and say okay all these three are 00:43:20 Yeshwanth Reddy Yerraguntla: Yes. Gopal Gottumukkala: related and the same. Now put all the artifacts came from all the three agents together and say these are the references but ultimately you need to go back and then change the event mechanism and then make it a direct service to service communication for now the I mean the the same kind of you know central core agent orchestrating different agent similar kind of philosophy is coming in all the areas I think probably that's where we end up also central agent having you know knowledge of multiple facets of the application depending on that it will pick its own rules and regulations and then proceed right so as of Yeshwanth Reddy Yerraguntla: All right. Gopal Gottumukkala: now whichever you Manisha Rasher you all together think the most practical and feasible path we should start there the next one will be the business Well, Yeshwanth Reddy Yerraguntla: H. Right. Gopal Gottumukkala: so why I'm saying first one should be very comfortable to all of us is because so that we don't deviate from our goals and thought process and the purity in the architecture. 00:44:46 Gopal Gottumukkala: Is there a possibility of us picking one or two flows now or you guys want to take time and Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: decide Yeshwanth Reddy Yerraguntla: Um, le I think I already have a couple of candidates. Van and Manisha can chip in and say which are important, which are not important. for me like the most critical thing is this uh pipeline because like I said everything is revolving around this guy Gopal Gottumukkala: Yeah. Yeshwanth Reddy Yerraguntla: only like if you don't show anything right to Gopal Gottumukkala: Ultimately, yes. Yeshwanth Reddy Yerraguntla: say that you have something on enterprise brain this even if you build everything else if this is not there it's still sort of like a weird this thing and this is sort of easy to implement also because we have done this so many times like this is one Gopal Gottumukkala: Yeah, Yeshwanth Reddy Yerraguntla: point pipeline where we are extremely confident on how to do it. So given the nature of Gopal Gottumukkala: then let's pick the time unless others have any Yeshwanth Reddy Yerraguntla: other yeah Gopal Gottumukkala: opinions. 00:46:00 Yeshwanth Reddy Yerraguntla: other so just this one Gopal Gottumukkala: What is going to happen in this pipeline and nothing fancy comes out because all of us have seen this many number of times. We don't feel excited about this. So we should be fine with that decision because what we are establishing is the platform itself. Satyasri Prabhakar Mantripragada: Okay. Gopal Gottumukkala: Guys, I mean Yeshwanth Reddy Yerraguntla: So adop question answer sir. Yeah please other people now. Gopal Gottumukkala: finish. Vara Kumar Jagarapu: Yeah, the thing is uh okay and uh I mean yeah to set the system what I feel is most important maybe next iteration we can concentrate on like Satyasri Prabhakar Mantripragada: It's Vara Kumar Jagarapu: I want to get you. Yeshwanth Reddy Yerraguntla: until Gopal Gottumukkala: Once this pipeline is established uh what we can do is one one side we can take the visualization part of it other Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: side alerting part of it. So that actually completes the platform. Everything else is all you know internal parts. Right? I'm just thinking out loud. 00:47:40 Satyasri Prabhakar Mantripragada: I don't Gopal Gottumukkala: I'm not saying that just to express my thought Satyasri Prabhakar Mantripragada: know. Gopal Gottumukkala: process. If I look from outside what I'm anticipating, I just want to sleep at home and then my phone should say go and uh talk to prick. there seems to be a change coming and once I get Satyasri Prabhakar Mantripragada: I'm going to take Gopal Gottumukkala: it then I go and uh you know Satyasri Prabhakar Mantripragada: one. Gopal Gottumukkala: understand more about the same thing then more or less this platform's job is done right so everything else is 20% use case it would have said something I wanted it to be told to me in a different way that's trained But that is like hardly 10% of the use case when I compare the number of users who are going to train right but that is critical so that comes next probably as long as uh you know Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: I'm getting the alerts and then the UI is dynamic. If I'm able to get the alert that inherently meaning that I'm able to understand I mean the system is able to understand what what needs to be presented to me right putting a UI layer that 00:48:58 Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: is the second one third one is anyway what we are starting at zero right that we I'm assuming that we solve that problem as long as a land there I should be able to talk to the system and understand more that completes the entire platform from a end user point of view obviously of course there Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: are 10 15 different layers that we need to build in architecture inside. But from outside, right? Yeshwanth Reddy Yerraguntla: Great. That is what you meant by shell, right? Gopal sir, Gopal Gottumukkala: Yeah. Yeah. Just want to Yes. Yeshwanth Reddy Yerraguntla: you can purge the rest of Gopal Gottumukkala: And in this three pieces I wanted to do the bare minimums wherever we know we can Yeshwanth Reddy Yerraguntla: it. Gopal Gottumukkala: validate the system do those bare minimums then application is built for me. How many sources it is touching or how scalable it is is the next level of uh priority for me. Yeshwanth Reddy Yerraguntla: Got Gopal Gottumukkala: Anyway, I know in the P I don't think correct me if I'm wrong. 00:50:07 Yeshwanth Reddy Yerraguntla: it. Gopal Gottumukkala: We are not going to prove the load. We are not going to prove the performance. Obviously, it has to be responsive. I'm not saying that you know after 2 minutes the result will come. What I'm talking about is system might not be able to answer 10 different questions at the same millisecond uh uh physical time and we are not going to put 200 data sources beneath right Satyasri Prabhakar Mantripragada: No, Yeshwanth Reddy Yerraguntla: Yeah. Satyasri Prabhakar Mantripragada: I'm saying we are not doing that. Gopal Gottumukkala: yeah I mean when we are not doing it I don't want to spend so much of time consuming one of these guys and then stabilize the platform. I would rather let the platform be flaky and stabilize the business aspects of it. Satyasri Prabhakar Mantripragada: Yeah, finally. So the platform needs to be solidified first uh so that we can build anything else on top of it. We don't have to come back and touch it unless there is any business requirement. So if we can get it to that level, uh we are 00:51:22 Gopal Gottumukkala: Yeah, basically paths the primary paths are well connected as we Satyasri Prabhakar Mantripragada: good. Gopal Gottumukkala: wanted and features can be keep on getting bulged on top of it. Satyasri Prabhakar Mantripragada: Yeah. So out of these uh we need to also identify uh which ones take multiple phases uh which ones can be done at one shot. For example, alerts and notifications. that can be done at one shot. You define the framework and hook it up to the respective systems and it is done. That is one implementation. Whereas there can be other Manisha Gundapuneedi: get some not sorry sir okay Yeshwanth Reddy Yerraguntla: Example, Satyasri Prabhakar Mantripragada: example whereas there can be Yeshwanth Reddy Yerraguntla: example. Manisha Gundapuneedi: fine Satyasri Prabhakar Mantripragada: other intelligence systems you need to build in incremental fashion in which case you need to define what is the MVP for now that we need to close the other aspects of intelligence we'll keep on adding in the next releases as in when there's a requirement or once we are done with the rest of the stuff then we'll go with that. 00:52:31 Satyasri Prabhakar Mantripragada: But at least the minimum level of functionality should be available in each of the listed items. Gopal Gottumukkala: Yes, Satyasri Prabhakar Mantripragada: We need to also find out what are the least most Gopal Gottumukkala: sir. Satyasri Prabhakar Mantripragada: dependent Gopal Gottumukkala: Yeah. Yeshwanth Reddy Yerraguntla: I want to add one point here. Uh praakar sir and gopal sir you put alert life cycle. If I take this as an example what we why is this moving? Sorry. what we build in first cut. If the expectation is it has to be perfect uh that we don't come back and touch it Gopal Gottumukkala: No no no no let me uh let me stop uh no no that's not the Yeshwanth Reddy Yerraguntla: again. Gopal Gottumukkala: point as well the behavior of it should be solidified that quality of that behavior will get enriched based on the knowledge that is getting built behind. Alerting seems like a framework from outside by the name of it but that also has an intelligence engine behind it right. So what Prabhakaru is meaning to say is assuming that the engine is giving something then we should 00:53:29 Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: have a solid path that how it behaves. So alert goes, a notification goes, email goes and dashboard there is something comes up that path how qualitative that alerting will be that will be Yeshwanth Reddy Yerraguntla: Got it. Gopal Gottumukkala: keeping improving but we don't want Yeshwanth Reddy Yerraguntla: qualitative like I understand that there will be an agent that will constantly Gopal Gottumukkala: to Yeshwanth Reddy Yerraguntla: change but I'm talking about certain things like um maybe we took a bad decision um while we came up with the architecture today that it has to change sometime maybe I'm like I don't know I'm talking something very basic only but changes That's all I'm trying to say. Gopal Gottumukkala: Okay. Yeah. Yeshwanth Reddy Yerraguntla: Even if you pull Gopal Gottumukkala: Absolutely. Yeshwanth Reddy Yerraguntla: something, Gopal Gottumukkala: See approach itself suggests that you are going to change your mind tomorrow. That's the whole point. Right. Yeshwanth Reddy Yerraguntla: right? Gopal Gottumukkala: The basic goodness of this approach is we are making the call today keeping the bigger goal Yeshwanth Reddy Yerraguntla: Exactly. 00:54:37 Gopal Gottumukkala: but more and more we touch the cross functionalities one of these things will get tinkered it has Yeshwanth Reddy Yerraguntla: Yes. Gopal Gottumukkala: to that's what we are not able to figure out right what are the cross dependencies we have between these Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: functionalities that's where we are struggling that's where the problem started right so we we are taking an approach thinking that this we can solidify and I mean put all the connectivity things could change. We have to be open for it. Yeshwanth Reddy Yerraguntla: Yes, Gopal Gottumukkala: So what I understood from his point of view is it is not like daily we'll come back and work on these alerts. Yeah, something drastically changes changes will come back and then fix it. Yeshwanth Reddy Yerraguntla: I see. Gopal Gottumukkala: It has to be fixed. That is what our next iteration, next iteration is going to do. But unless and until that necessity comes, the alert will work. Two things constantly has to happen is whatever we did till yesterday is in shape or not. If it is not in shape, we have to correct it and then that corrections are not going to be permanent in the sense we don't know maybe whatever we built today on uh uh adha question and answering life cycle right it could only sustain for a week after that there might be some changes coming in we should be open to do those changes and then move forward again that second set of changes also not going to be permanent But luckily they permanent 00:56:13 Yeshwanth Reddy Yerraguntla: that I'm we are on the same page in that Gopal Gottumukkala: also. Yeshwanth Reddy Yerraguntla: case constant like there is no perfection is always Gopal Gottumukkala: Yeah. Yeshwanth Reddy Yerraguntla: sort of a local phenomena immediate requirement but overall vision but we have to come up with the best call uh so that it is available to changes tomorrow and at the same time we are doing our best based on the knowledge that we have both on what to do which people to u satisfy and what overall vision whatever All we are taking it's always the local optimum and that is good enough to Gopal Gottumukkala: Yes sir. Yeah. Yeshwanth Reddy Yerraguntla: know Gopal Gottumukkala: Uh uh I mean at least the current situation I don't see any other way unless and until somebody can help us devise it better. So, my intention is to get to the Yeshwanth Reddy Yerraguntla: I agree agree Gopal Gottumukkala: floor. Yeshwanth Reddy Yerraguntla: I was reading this book recently. Uh Eric Ran startup and there he also tells very similar things only that you if you think you are going to build something that you can show 6 months down the line or even 1 month down the line you're already in the wrong approach. 00:57:52 Yeshwanth Reddy Yerraguntla: What you have to show is something that you have to think through and try to build it as soon as possible with the right set of compromises. Clarity line up that's all we can do. Yeah. Like instead of just sitting and thinking what all cross functionalities will be there Gopal Gottumukkala: I strongly believe it will come at clarity will come once we start we'll get clarity Yeshwanth Reddy Yerraguntla: especially Gopal Gottumukkala: only thing why I'm insisting so many times is suddenly we should not feel that you know we are building throw away building throw away it is not the case. We are keep on correcting it. As long as we do that, we should be fine. Otherwise in a week's time, we will still question ourself that are we doing the right thing or not. We have to be patient enough for at least couple of weeks. Once we get the two weeks done, I strongly believe that we'll have clarity on end to end picture in in code. what I'm saying anyway we have clarity theoretical clarity we already have but we'll get much better uh clarity and we can actually match to the vision also in two 00:59:06 Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: weeks time the layers will form connections will establish then probably you'll have a number of people who can derive the Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: low-level tasks Yeshwanth Reddy Yerraguntla: Understood. Actually speaking of that point um like if we okay first thing I understood is let's have this uh priority uh matrix on what are all the things that we think are super duper important what which can be postponed based on business requirements technical feasibility and uh overall architecture uh importance That said, let's say we have decided these three items are P 0 or these four items. Next question I'll have is um how do we ensure this is translated to a proper work job both in Jira as well as for the juniors who want to build it because there there are two options at every uh item right either it is super clear and super known or it is unknown it's Gopal Gottumukkala: Thank Yeshwanth Reddy Yerraguntla: a it's a it's not a binary exactly but on a spectrum either the task is very 01:00:29 Gopal Gottumukkala: you. Yeshwanth Reddy Yerraguntla: easy to implement and very clear or it's not easy and not clear our situation slow do you suggest uh we put hold on the junior saying you wait for us to finish the P 0 tasks so that we have clarity on what to do at P1 or uh somehow we will tell them okay these are the acceptance criteria you go figure out what to do or it has to be a hybrid of somewhere instructions. I want to I also want to understand like that was the second agenda Gopal Gottumukkala: So first cut we pick the core Yeshwanth Reddy Yerraguntla: item. Gopal Gottumukkala: and anyway there are certain things we are solid that we some some task needs to be done in that P usually what happens is we will try to do they I mean those Yeshwanth Reddy Yerraguntla: H. Gopal Gottumukkala: tasks also If if we know there are certain things that we'll pick it up in 3 days time not now but we we'll have to do in order to finish this P in the within this P there will be task very independent but it needs to be 01:01:45 Yeshwanth Reddy Yerraguntla: M. Gopal Gottumukkala: connected connectivity part will take care let the task be done by the folks because you reach to a connectivity in two days standard they can work on that uh imple implementation of independent things Yeshwanth Reddy Yerraguntla: Mhm. Gopal Gottumukkala: then primary folks can actually do that orchestration connectivity. Yeshwanth Reddy Yerraguntla: H Gopal Gottumukkala: So that way you can engage the team from right now and they'll also get contested. So it'll be a little you know uh they might not have the complete picture of what is being built but that's okay. When you actually connect things they just need to understand how things are connected. Everything else they already know because they only did it. Otherwise, TT becomes too wide 624 block numbers third. Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: So visual visualization part is solid once you Yeshwanth Reddy Yerraguntla: Right. Gopal Gottumukkala: decided representation is pretty straightforward right as a chart interface there is not not much of magic and then similarly in the chart interface figuring out sockets and stream is pretty standard technique though it is it is complex and rich but all the developers should be able to do it right so take off all those things, put them in in any order, 01:03:23 Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: hand it over to the people that are available. They will solve those problems which you don't need to sit and then figure out, right? Even the library to be used, anything that they can figure out. If you're not happy with those libraries, Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: we can always go back and work with a different library. But be one would have been laid out. in the in the same A. Yeshwanth Reddy Yerraguntla: Understood. Yeah. Break the whole task itself into knowns and unknowns. Let the seniors handle unknowns. Gopal Gottumukkala: Uh, Yeshwanth Reddy Yerraguntla: Let the juniors handle known aspects. Gopal Gottumukkala: no. Yes. Yeshwanth Reddy Yerraguntla: Yeah. Okay. So in that case the next item for every um this thing will be to break them known unknown clarity on we know what to do makes sense. Gopal Gottumukkala: Thank Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: you. Yeshwanth Reddy Yerraguntla: Um what I can do is Manisha uh V myself Manisha will connect and we'll try to discuss which are P 0 and which are P1's uh if Prabhakasar is available he can also uh connect with us u based on the P 0 we'll know what needs to be 01:05:19 Gopal Gottumukkala: Yeah, suggest there is a there is a known with 20% unknown. Yeshwanth Reddy Yerraguntla: Okay. Okay. Okay. Gopal Gottumukkala: Every damn thing that we are trying to do there is a certain un unknown part of it. Right? Yeshwanth Reddy Yerraguntla: Right. Gopal Gottumukkala: So if some task which has about 60 70% 60 I still give Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: that 60 70 to somebody else. Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: because there is unknown I don't want to keep the Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: taskown. Yeshwanth Reddy Yerraguntla: Okay. Okay. Okay. Gopal Gottumukkala: So, Yeshwanth Reddy Yerraguntla: What if the unknown I don't know maybe this is really stupid question I'm about to Gopal Gottumukkala: okay. Yeshwanth Reddy Yerraguntla: ask. What if the unknown is a dependency on the known aspect like uh uh Mhm. Gopal Gottumukkala: Most of the times you can assume as Yeshwanth Reddy Yerraguntla: Yeah. Gopal Gottumukkala: most of the Yeshwanth Reddy Yerraguntla: Oh, okay. Got it. Gopal Gottumukkala: times Yeshwanth Reddy Yerraguntla: Mark Gopal Gottumukkala: they'll struggle. 01:06:59 Yeshwanth Reddy Yerraguntla: there. Yeah. Gopal Gottumukkala: So anything that is solid hand it over to them even if it is half and half it is still fine because we have the maturity of pulling that uh 50% then we are done with the remaining you will be able to create Yeshwanth Reddy Yerraguntla: H Gopal Gottumukkala: enough bucket for Yeshwanth Reddy Yerraguntla: yeah, that's what I was struggling with. Gopal Gottumukkala: that risk we have to take already at least I don't know even it could be one day it is still worth So make an attempt and if nothing is coming, Yeshwanth Reddy Yerraguntla: Okay. Gopal Gottumukkala: nothing is coming. But our mindset should be very clear that anything that is known, we should not be making an effort. Yeshwanth Reddy Yerraguntla: H Gopal Gottumukkala: Something will fall into everybody's Yeshwanth Reddy Yerraguntla: Makes Gopal Gottumukkala: bucket. Yeshwanth Reddy Yerraguntla: sense. We'll give it an attempt or not. Gopal Gottumukkala: I'm sure I'm pretty confident enough bucket will come for anybody can do kind Yeshwanth Reddy Yerraguntla: Yeah. Yes. Gopal Gottumukkala: of Yeshwanth Reddy Yerraguntla: Yeah. 01:08:34 Yeshwanth Reddy Yerraguntla: We'll try. Shall we continue in this call or what do you suggest? Vara Kumar Jagarapu: share the Yeshwanth Reddy Yerraguntla: Okay. Gopal sir. Vara Kumar Jagarapu: number. Yeshwanth Reddy Yerraguntla: Um. Ah. Okay. Gopal Gottumukkala: But call me nowadays. Yeshwanth Reddy Yerraguntla: Sure Gopal Gottumukkala: I'm not sitting on the machine for long time between just call Yeshwanth Reddy Yerraguntla: sir. Gopal Gottumukkala: me. Yeshwanth Reddy Yerraguntla: Sorry sir. Gopal Gottumukkala: Thanks. Yeshwanth Reddy Yerraguntla: Yeah, thank you so much Satyasri Prabhakar Mantripragada: and kiss them. Vara Kumar Jagarapu: Thank you, Yeshwanth Reddy Yerraguntla: sir. Vara Kumar Jagarapu: sir. Thank you. Yeshwanth Reddy Yerraguntla: 10 minutes. Satyasri Prabhakar Mantripragada: I'll stay back and uh if you Yeshwanth Reddy Yerraguntla: Okay. Satyasri Prabhakar Mantripragada: are Yeshwanth Reddy Yerraguntla: Yeah. Yeah. Satyasri Prabhakar Mantripragada: prioritizing Yeshwanth Reddy Yerraguntla: Um, Satyasri Prabhakar Mantripragada: how are you going on the sheet anyways? Yeshwanth Reddy Yerraguntla: yeah. Yeah. This s*** only, right? Yeah. Um, I mean both are sort of Satyasri Prabhakar Mantripragada: High level to do to 01:09:56 Yeshwanth Reddy Yerraguntla: overlaps. Agree. Agree. Yeah. Satyasri Prabhakar Mantripragada: do Yeshwanth Reddy Yerraguntla: Let me just do a P 0 on this uh the way I will I don't know on oh how Satyasri Prabhakar Mantripragada: on high level tasks, huh? Uh high level task. Yeshwanth Reddy Yerraguntla: P1 high level Satyasri Prabhakar Mantripragada: Okay. Yeshwanth Reddy Yerraguntla: tasks Satyasri Prabhakar Mantripragada: No, no, you're you're on the right one. So I was only to high level tasks, Yeshwanth Reddy Yerraguntla: H. Satyasri Prabhakar Mantripragada: but that's Yeshwanth Reddy Yerraguntla: Oh, this one. Satyasri Prabhakar Mantripragada: fine. Yeshwanth Reddy Yerraguntla: Oh, actually, yeah. Same thing, Ana. It looks like it's the same thing. Satyasri Prabhakar Mantripragada: One or two levels up. Yeshwanth Reddy Yerraguntla: Right, right, right. Yeah. Satyasri Prabhakar Mantripragada: We should start Yeshwanth Reddy Yerraguntla: All right. All choose. Satyasri Prabhakar Mantripragada: working. Yeshwanth Reddy Yerraguntla: Yeah, I mean this is definitely getting covered in this uh list because Oracle Unifier. Okay, that is one exception where I didn't write anything specifically about it. 01:11:10 Yeshwanth Reddy Yerraguntla: But uh when we are dealing with uh okay block numbers just one second I know what to do. component. So first few are UI components alert designs in black. Satyasri Prabhakar Mantripragada: What is Vara Kumar Jagarapu: She already Satyasri Prabhakar Mantripragada: happening? Vara Kumar Jagarapu: mentioned story breakdown Yeshwanth Reddy Yerraguntla: Yeah. Okay. Right. Right. Right. Yeah. Vara Kumar Jagarapu: category. Yeshwanth Reddy Yerraguntla: Okay. Oh, same thing. Sorry. Data source agent law. There are a couple of P 0 which directly translate to unifier tasks. So uh praa cursor like if Satyasri Prabhakar Mantripragada: Sorry. Data source data Yeshwanth Reddy Yerraguntla: I if I take a Satyasri Prabhakar Mantripragada: ship. Yeshwanth Reddy Yerraguntla: second if I take this as Satyasri Prabhakar Mantripragada: Sure. Yeshwanth Reddy Yerraguntla: category clear data source agent these these are these are specific to unifier or any other data source agent right we are saying these are P 01:12:47 Satyasri Prabhakar Mantripragada: Mhm. Yeshwanth Reddy Yerraguntla: 0 which which translate to that unifier agent Satyasri Prabhakar Mantripragada: Yes. But correct. Yeshwanth Reddy Yerraguntla: anyway Satyasri Prabhakar Mantripragada: But as of now probably we can hold on unifier till we get that NDA thing and all that sorted out. We'll focus on the core platform. Yeshwanth Reddy Yerraguntla: okay Satyasri Prabhakar Mantripragada: Then we'll move on to integrations. Yeshwanth Reddy Yerraguntla: actually yeah P 0 P There was a circuit. Oh, Manisha Gundapuneedi: Fore Yeshwanth Reddy Yerraguntla: you Satyasri Prabhakar Mantripragada: one we may not focus exactly on unifier but we may have to build the Manisha Gundapuneedi: connectors Satyasri Prabhakar Mantripragada: underlying foundation to connect to it. Manisha Gundapuneedi: framework. Satyasri Prabhakar Mantripragada: The connect Manisha Gundapuneedi: connector Yeshwanth Reddy Yerraguntla: a connector. Manisha Gundapuneedi: framework. Yeshwanth Reddy Yerraguntla: That by itself is not something which will give us any business value or you have the ability to build other agents. Correct. Manisha Gundapuneedi: As in scalable way other connection other connector Yeshwanth Reddy Yerraguntla: support you are simply building a starting point for building other connectors. 01:14:33 Yeshwanth Reddy Yerraguntla: If you have a way of saying okay this connector is API this connector is rest based this connector is so and so based you can initialize connectors in that way now secondw Manisha Gundapuneedi: Carta. Yeshwanth Reddy Yerraguntla: which is not correct. So we have to say only maximum one one and a half item perh category should be P 0. So you can like discuss and change these uh items from P 0 to P1 P2 P3. I'll just come back. Manisha Gundapuneedi: question. Enter the only Vara Kumar Jagarapu: So then I think we can directly keep it. Uh yeah, let me share my screen. Yeah, my screen can be sh Yeah. So all these things okay so now going by this components chart interface key what what's the reality Okay. Satyasri Prabhakar Mantripragada: breaking down or if you're giving a priority uh what is the parent link for this if chat interface becomes P 0 or P1 or P2 what is the parent that we are trying to accomplish from the highle tasks 01:17:04 Vara Kumar Jagarapu: Now that Satyasri Prabhakar Mantripragada: we may have say five stories or three stories or two stories Please second. Vara Kumar Jagarapu: Okay. So yeah one cycle rate. So initial pictures like which we can complete to get more stabilized version to add on things. So what? Hello. Satyasri Prabhakar Mantripragada: We're done. Vara Kumar Jagarapu: Uh so when it comes to P1 on top of it let me show this already while we are in call Satyasri Prabhakar Mantripragada: Okay. Vara Kumar Jagarapu: with Gopal. So this is already Satyasri Prabhakar Mantripragada: Thank you. Vara Kumar Jagarapu: PJ is not driven from the business present. Satyasri Prabhakar Mantripragada: I don't know. So first We have some Vara Kumar Jagarapu: Manisha. So Manisha Gundapuneedi: voice. Vara Kumar Jagarapu: new voice I'm addable Satyasri Prabhakar Mantripragada: Now the break of the low and high Vara Kumar Jagarapu: now. Manisha Gundapuneedi: Nada. Satyasri Prabhakar Mantripragada: low. Manisha Gundapuneedi: Nada. Vara Kumar Jagarapu: Let me re log in again. Manisha Gundapuneedi: I don't 01:24:03 Vara Kumar Jagarapu: Okay. So I can share my screen. Okay. Manisha Gundapuneedi: Ah, Vara Kumar Jagarapu: P it is related to life. Manisha Gundapuneedi: Pete. Vara Kumar Jagarapu: If you want to complete Chart interface question answer data sources transport. Okay, let's Manisha Gundapuneedi: Hopefully. Vara Kumar Jagarapu: see. Manisha Gundapuneedi: Okay. Vara Kumar Jagarapu: This is page zero. Let me filter for T1. So P1C context circle Yeshwanth Reddy Yerraguntla: I am Vara Kumar Jagarapu: 20 seconds. Okay. Yeshwanth Reddy Yerraguntla: back. Vara Kumar Jagarapu: Yeah, already present. Yeshwanth Reddy Yerraguntla: Oh, you Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: just Vara Kumar Jagarapu: Yeah. more important Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: on top of it. P1 P2 that is also P right discuss P1 P2 Yeshwanth Reddy Yerraguntla: Are we discuss? Just shut down P 0 and Vara Kumar Jagarapu: P thesis P having any other Yeshwanth Reddy Yerraguntla: Okay. Vara Kumar Jagarapu: opinions maybe kind of Q&A. Life user interfaces. 01:28:09 Yeshwanth Reddy Yerraguntla: Uh near. Vara Kumar Jagarapu: This is already covering covering Yeshwanth Reddy Yerraguntla: Yes. Yes. Yes. Vara Kumar Jagarapu: This this is also going to cover as part of Yeshwanth Reddy Yerraguntla: said, Vara Kumar Jagarapu: corres. Okay. Yeah. So maybe Yeshwanth Reddy Yerraguntla: "Okay. Vara Kumar Jagarapu: next Chart history. Continuous conversations Yeshwanth Reddy Yerraguntla: Oh, Vara Kumar Jagarapu: authentication. Yeshwanth Reddy Yerraguntla: okay. Authentication. I think P. Why I'm saying P 0 is because history conversations. Anyway, we have implemented it. So, we know how to do it. So, better to treat it as P 0. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: Next. You sure? Vara Kumar Jagarapu: You send the dashboard Yeshwanth Reddy Yerraguntla: Yeah. Core agent. Vara Kumar Jagarapu: this news from the core agent for which user Yeshwanth Reddy Yerraguntla: Uh Vara Kumar Jagarapu: They could be Yeshwanth Reddy Yerraguntla: either Vara Kumar Jagarapu: much Yeshwanth Reddy Yerraguntla: If there are okay if there are 20 users 20 users need to get news in a different uh fashion right that is the core uh functionality of this item. 01:30:50 Yeshwanth Reddy Yerraguntla: Um, Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: should we really call it P1 or P2 is the question. P1. How much can we do in uh two weeks is the next question. Two weeks. We'll probably have to take shortcuts. Manisha Gundapuneedi: Fore Yeshwanth Reddy Yerraguntla: got over here. Manisha Gundapuneedi: record. Yeshwanth Reddy Yerraguntla: I mean from a business perspective this is definitely a key feature. I I will uh that for with that for that reason I will say this is P1 but when it comes to implementation uh how much can we do in one or two weeks like even if it is P1 what I'm saying is there are a lot of unknowns in that um a lot and I would say Almost 90 10% on 90% unknown Vara Kumar Jagarapu: Okay. Two weeks. Yeshwanth Reddy Yerraguntla: started. Vara Kumar Jagarapu: Outcome point of Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: view. Yeshwanth Reddy Yerraguntla: Outcome point of view law two weeks slow. Um see the fact that I have to deliver news in a customized fashion means first of all I have to create news for the organization store news I have to update news I have to refresh update refresh of tell news after doing these things I need 01:32:48 Vara Kumar Jagarapu: Hello. Yeshwanth Reddy Yerraguntla: to have another agent which will know how to distribute news to every person. Uh each one of these things are unknowns right now. Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: Until we finish these four, we will not know how to demonstrate Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: this. Vara Kumar Jagarapu: And we go to agreed upon initial not going to take complex outcomes dash there should be some dashboard. So working version or stable version Yeshwanth Reddy Yerraguntla: Yes. Vara Kumar Jagarapu: Next. Yeshwanth Reddy Yerraguntla: I feel it should be paid to. Vara Kumar Jagarapu: Uh yeah. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: Training life cycle. Training related P3, right? P3. Yeshwanth Reddy Yerraguntla: Okay. P2 P3 we'll have to clearly Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: agree to one Vara Kumar Jagarapu: Okay. So, Yeshwanth Reddy Yerraguntla: thing. Vara Kumar Jagarapu: let me Yeshwanth Reddy Yerraguntla: So why I thought P3 was this like who is this copaler was saying that this product is unusable and that's not the case without training life cycle also a lot of uh uh 01:34:52 Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: usage can happen on the Vara Kumar Jagarapu: But the Yeshwanth Reddy Yerraguntla: platform. Vara Kumar Jagarapu: valuepect if you are giving instruction that is respecting Uh Yeshwanth Reddy Yerraguntla: Yes, true. That is that's definitely true. Vara Kumar Jagarapu: that's Yeshwanth Reddy Yerraguntla: Um, let's just pause here. So, the question is whether it should be P2 or P3 or P1, right? Vara Kumar Jagarapu: yeah P 0 already which is related to Q&A life cycle P1 P2 P3. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: Okay. So as a next loop Yeshwanth Reddy Yerraguntla: Right. Vara Kumar Jagarapu: with outcome Yeshwanth Reddy Yerraguntla: Yeah. Yeah. Vara Kumar Jagarapu: right Yeshwanth Reddy Yerraguntla: So you you have filtered everything except P 0, right? Okay. Vara Kumar Jagarapu: yes Yeshwanth Reddy Yerraguntla: Blanks. Uh keep them in case we missed anything. Vara Kumar Jagarapu: and give Yeshwanth Reddy Yerraguntla: Uh yeah, Vara Kumar Jagarapu: it Yeshwanth Reddy Yerraguntla: I think there is only one blank at the very bottom which is designs are the Vara Kumar Jagarapu: Yes. 01:36:09 Vara Kumar Jagarapu: Yes. Not Yeshwanth Reddy Yerraguntla: P. Oh, LLM tracing. Vara Kumar Jagarapu: this. Yeshwanth Reddy Yerraguntla: Okay. Okay. Monitoring per user P. Yeah. Vara Kumar Jagarapu: Yeah. Show them. No. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: Okay. So, news and insights. It is dashboard. P1 P2 training life cycle. Okay. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: I'll go for another life second. Yeshwanth Reddy Yerraguntla: Alerting. See I I am thinking of priority in two different dimensions. This is for both all the people, Manisha and V. One is how easy is it to implement two how much business impact it is bringing. These are the two dimensions using which we want to take up priority, right? Because if something is super critical for business but it is extremely hard to implement, Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: we will only waste time. Conversely, if it is super easy to implement but business really doesn't care, again we are wasting our time to some extent. 01:37:27 Yeshwanth Reddy Yerraguntla: So we have to strike a balance where things are easy and also drive high business value P Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: 0 we are clear that it might be somewhat medium difficulty in implementation but we can't do anything else without it. So we don't have a choice. P1 on the other hand we want to impress as many business people as we can uh while trying to ensure that we are not Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: really burning uh our uh uh our head our brain our hands anything we are not burning our energy. Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: So uh how to come up with P1? One thing is that showing of news is a high business uh item. Training also to to an extent I agree with you that it is a high business uh priority. Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: Training to an extent is not that difficult to implement. Also we've been implementing it to some extent. Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: like Amulia was able to do it uh reasonably well over the last couple of weeks. 01:38:46 Yeshwanth Reddy Yerraguntla: So we know it's not super difficult. So if you say we want to take it up as P1, Vara Kumar Jagarapu: Peace. Yeshwanth Reddy Yerraguntla: I'm I'm okay with it. Alert life cycle on the other hand also high business value high easy to implement. Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: So they go to P1 pet code. So training training also you can put as Vara Kumar Jagarapu: P1. Yeshwanth Reddy Yerraguntla: P1. Vara Kumar Jagarapu: Okay. Can have good less time more Yeshwanth Reddy Yerraguntla: Good. Vara Kumar Jagarapu: clear. Yeah. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: So data visualization like we present P1 we can do this Yeshwanth Reddy Yerraguntla: Yeah. Next uh activity is uh generating news life Vara Kumar Jagarapu: generate generate new Yeshwanth Reddy Yerraguntla: cycle. Right. So we have data agents real time. Vara Kumar Jagarapu: left. Yeshwanth Reddy Yerraguntla: This is a P1 in terms of business. Definitely this is high priority but it is equally very difficult to implement. 01:40:01 Yeshwanth Reddy Yerraguntla: So we will have to probably break this down Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: into uh a P1 just that we will not be able to deliver an output. This is what I'm thinking. Vara Kumar Jagarapu: initially. Yeah. Yeah, I agree. Maybe. Yeah. Small parts of the things needs to be completed which is critical. Yeshwanth Reddy Yerraguntla: We can't wait for this thing because this be a large Vara Kumar Jagarapu: Oh Yeshwanth Reddy Yerraguntla: activity. Vara Kumar Jagarapu: snapshot is still there. Yeshwanth Reddy Yerraguntla: and and secondly this item has a lot of Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: um other items depend on this item very much. User broadcast awareness agent is waiting on this guy. Data snapshot uh needs this guy. Um user a visualization agent also that to show things on dashboard in a personalized way that is also dependent on Vara Kumar Jagarapu: Mhm. Yeshwanth Reddy Yerraguntla: this guy. So even though we will not be able to deliver on the user interface anything we can't afford to 01:41:20 Vara Kumar Jagarapu: So then then it can be P Yeshwanth Reddy Yerraguntla: wait it is Vara Kumar Jagarapu: 0. Yeshwanth Reddy Yerraguntla: definitely parallel activity the high research intense activity which Vara Kumar Jagarapu: It will not be delivered initially but part by Yeshwanth Reddy Yerraguntla: yeah Vara Kumar Jagarapu: part I be zero then Yeshwanth Reddy Yerraguntla: Right. I I completely am aligned with you. Vara Kumar Jagarapu: but Yeshwanth Reddy Yerraguntla: Some research like whether it is me or Raj or Manisha or start after two weeks something will be able to show that there is some progress. Vara Kumar Jagarapu: Yeah. Then then it's fine. I kept it as P Yeshwanth Reddy Yerraguntla: What I can suggest is Oh, Vara Kumar Jagarapu: with just Yeshwanth Reddy Yerraguntla: so we can break it down into parallel tracks. Vara Kumar Jagarapu: another. Yeshwanth Reddy Yerraguntla: P 0.10.0.3 basically every one of them will be a parallel track eventually. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: Oh, okay. Yeah. Vara Kumar Jagarapu: step. Yeshwanth Reddy Yerraguntla: We agree that this is P 0. Once this is finished then we are saying uh what is line number oh P 01:43:14 Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: 0.12 Vara Kumar Jagarapu: connection can establish. Yeshwanth Reddy Yerraguntla: there let me just think it through I didn't realize what you were saying receive all updates from data source agents per refresh fetch or Vara Kumar Jagarapu: Uh Yeshwanth Reddy Yerraguntla: snapshot or snapshot intelligence broadcast news to user aware in insight engine. So each of these is a different item. Vara Kumar Jagarapu: I don't think Yeshwanth Reddy Yerraguntla: 0.1 is like receive all updates from all data associations refresh detailed. It's not so Vara Kumar Jagarapu: Uh critical part Yeshwanth Reddy Yerraguntla: simple. Vara Kumar Jagarapu: actually Yeshwanth Reddy Yerraguntla: Oh no, no. Yeah, Vara Kumar Jagarapu: previous kind of Yeshwanth Reddy Yerraguntla: they are. Yeah, Vara Kumar Jagarapu: integration Yeshwanth Reddy Yerraguntla: one and three are straightforward unknown. You're right. Unknown, but still some 8020 unknowns. First, third, second is like 100% unknown. Vara Kumar Jagarapu: So Yeshwanth Reddy Yerraguntla: Yeah. Yeah, Vara Kumar Jagarapu: dependencies. So, Yeshwanth Reddy Yerraguntla: correct. Vara Kumar Jagarapu: okay. 01:44:42 Vara Kumar Jagarapu: Already think the Yeshwanth Reddy Yerraguntla: Yeah, Vara Kumar Jagarapu: Okay, next broadcast Yeshwanth Reddy Yerraguntla: broadcasting news. So given once you receive the news, Vara Kumar Jagarapu: news. Yeshwanth Reddy Yerraguntla: you have to tell it to everyone. This will become P1 or P2 and it is dependent on this item. Okay. Vara Kumar Jagarapu: it Yeshwanth Reddy Yerraguntla: P1. Vara Kumar Jagarapu: actually Yeshwanth Reddy Yerraguntla: What are you saying? Vara Kumar Jagarapu: uh nothing uh chart interface which is P right we can see Yeshwanth Reddy Yerraguntla: Yes. Vara Kumar Jagarapu: it uh Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: so dashboard We kept it inside Yeshwanth Reddy Yerraguntla: Uhhuh. Vara Kumar Jagarapu: P2. Yeshwanth Reddy Yerraguntla: Dashboard. Uh yeah to we have to take that decision dashboard. Vara Kumar Jagarapu: And Yeshwanth Reddy Yerraguntla: Will they use it or the dashboard by the end of everything will become P1. So you have to dearize. Yeah. Correct. Correct. Yeah. Vara Kumar Jagarapu: dashboard. At 01:46:21 Yeshwanth Reddy Yerraguntla: Makes sense. Makes sense. Vara Kumar Jagarapu: least Yeshwanth Reddy Yerraguntla: 100% uh bandwidth. However that said if you take a par tracks which is like Manisha sorry Manisha Ganesha say they are responsible for UI tasks once they finish the chat aspects which is P 0 they don't have anything on P1 what I'm Vara Kumar Jagarapu: Hit Yeshwanth Reddy Yerraguntla: trying to say is once they deliver P 0 what What are they doing? Vara Kumar Jagarapu: now list at the same time Yeshwanth Reddy Yerraguntla: What's the Vara Kumar Jagarapu: three things. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: So for this particular P like actually we having we are having three tasks which can based on the priority we can assign to Yeshwanth Reddy Yerraguntla: Yes. Vara Kumar Jagarapu: them. life cycle those tasks can be assigned right. Yeshwanth Reddy Yerraguntla: Got it. Got Vara Kumar Jagarapu: Uh so so whatever it may be P 0 or P1 or Yeshwanth Reddy Yerraguntla: it. Vara Kumar Jagarapu: P2 Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: independent 01:48:03 Yeshwanth Reddy Yerraguntla: P1 makes sense. Vara Kumar Jagarapu: Right. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: Outcome point of I think P already agreed upon you put P1. Okay. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: These are the certain features less complexity we can deliver it. These are more straightforward. We can do that. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: When it is P2 we'll be doing parallelly based on the bandwidth Yeshwanth Reddy Yerraguntla: We'll have to start early. Vara Kumar Jagarapu: we have. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: We we have to start it early. But Yeshwanth Reddy Yerraguntla: Got it. Vara Kumar Jagarapu: but this needs to be start when if these people are having Yeshwanth Reddy Yerraguntla: Got it. Vara Kumar Jagarapu: bandwidth. Yeshwanth Reddy Yerraguntla: Yes. Vara Kumar Jagarapu: We are already occupied with this Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: delivery. Yeshwanth Reddy Yerraguntla: Correct. Margin. Makes sense. Makes sense. Vara Kumar Jagarapu: So I think I think point of view we Yeshwanth Reddy Yerraguntla: Yeah, Vara Kumar Jagarapu: can 01:49:20 Yeshwanth Reddy Yerraguntla: we can say that this is P2. Vara Kumar Jagarapu: Oh. Satyasri Prabhakar Mantripragada: I missed the conversation. Yeshwanth Reddy Yerraguntla: Yeah. Satyasri Prabhakar Mantripragada: I was looking into something else. Yeshwanth Reddy Yerraguntla: Summary is the answer. P 0 we have clarity that it is all about creating the right um foundation for the Satyasri Prabhakar Mantripragada: Huh? Yeshwanth Reddy Yerraguntla: whole project. of which the one feature that we will be able to demonstrate is the chat interface itself which we anyway have in enterprise brain kakapote the change here is going to be it's a fresh architecture so we'll build again from beginning um next big change is the interface will be react based instead of uh whatever the pyog based right Satyasri Prabhakar Mantripragada: Okay. If we have to restart from scratch, Yeshwanth Reddy Yerraguntla: Um Satyasri Prabhakar Mantripragada: are we only restarting the front end part of it or the Yeshwanth Reddy Yerraguntla: only the front look only the front and back end we know how to do it. Satyasri Prabhakar Mantripragada: entire Okay. Yeshwanth Reddy Yerraguntla: We will we will continue use it. 01:50:18 Satyasri Prabhakar Mantripragada: And how much is the effort? How much is going to be the effort Yeshwanth Reddy Yerraguntla: How much uh the load like Satyasri Prabhakar Mantripragada: there time Yeshwanth Reddy Yerraguntla: time wise Satyasri Prabhakar Mantripragada: wise? Yeshwanth Reddy Yerraguntla: antara Satyasri Prabhakar Mantripragada: And if it is 1 2 hours maybe 1 2 hours or 1 day should be good enough on actually Yeshwanth Reddy Yerraguntla: Mhm. Vara Kumar Jagarapu: UI chart Satyasri Prabhakar Mantripragada: m Vara Kumar Jagarapu: interface and questions and answers UI point of view and core point of data agent point of these are the we thought of picking and completing. Yeshwanth Reddy Yerraguntla: So from a P 0 delivery perspective, back end I can surely tell we can deliver it by next uh Vara Kumar Jagarapu: So Yeshwanth Reddy Yerraguntla: Monday. Satyasri Prabhakar Mantripragada: Thank you. Yeshwanth Reddy Yerraguntla: front end side I don't know react I don't know the technology behind copilot kit so either Manisha or Vara Kumar Jagarapu: heat Yeshwanth Reddy Yerraguntla: wankatana should take that call only Satyasri Prabhakar Mantripragada: discussed and Manisha find out what needs to be done. Yeshwanth Reddy Yerraguntla: Yeah, Raakar sir. 01:51:56 Yeshwanth Reddy Yerraguntla: One thing I was saying is the there is one uh known in the P 0 sorry in the UI part which is because we have done so much with py log we know how to do certain things and how to show flowcharts how to Satyasri Prabhakar Mantripragada: H. Yeshwanth Reddy Yerraguntla: show what's it called how to show graphs we to some extent the technology we know how to do it but through react how to do it is the unknown. So definitely I would want Ara or Manisha or Rajshakhar to uh commit to a a number given the number of people who are working is only two people which is Man Syanita and Ganesh and it is being led by Manisha. Satyasri Prabhakar Mantripragada: Mhm. Handshake between Pylo and React React Yeshwanth Reddy Yerraguntla: Mhm. Satyasri Prabhakar Mantripragada: renders. Yeshwanth Reddy Yerraguntla: uh pylog itself is the rendering framework. So and we are completely replacing it with react. Satyasri Prabhakar Mantripragada: Okay. Yeshwanth Reddy Yerraguntla: The only handshake that happens is from the agent. Satyasri Prabhakar Mantripragada: Okay. Yeshwanth Reddy Yerraguntla: Now pylog used to take agents uh messages and transform it on the UI. 01:53:12 Yeshwanth Reddy Yerraguntla: Now react has to do do that drop in replacement onto Satyasri Prabhakar Mantripragada: Okay. Vara Kumar Jagarapu: Yeah, sorry. Exploration maybe then summary. Satyasri Prabhakar Mantripragada: What Yeshwanth Reddy Yerraguntla: Uh, Satyasri Prabhakar Mantripragada: is Vara Kumar Jagarapu: Okay, I think some person is doing that type what I got to know Yeshwanth Reddy Yerraguntla: knock Vara Kumar Jagarapu: from either Okay. Estimations. Yeah. Then maybe Manisha Gundapuneedi: Okay. Vara Kumar Jagarapu: call this coach. Manisha Gundapuneedi: Okay. Satyasri Prabhakar Mantripragada: That's great. Vara Kumar Jagarapu: Yeah. Yeah. Yeshwanth Reddy Yerraguntla: Uh just one Vara Kumar Jagarapu: Sure. Satyasri Prabhakar Mantripragada: organization. Yeshwanth Reddy Yerraguntla: thing. I group. Okay. Yeah. Uh so sorry summary P 0 is that P1 Satyasri Prabhakar Mantripragada: Here Yeshwanth Reddy Yerraguntla: is Braaser. P 0. Everything else we want to now prioritize as P1, P2, P3. And what we are thinking is P1 I mean this is our uh uh our logic. P1 is high business value, easy to implement. 01:55:20 Yeshwanth Reddy Yerraguntla: P2 is high business value but uh difficult to implement. Satyasri Prabhakar Mantripragada: we go. Yeshwanth Reddy Yerraguntla: so that we'll take some more time to parallelly kickstart those activities while P1 happens. P3 on the other hand is like low business value and unknown level of implementation. It might be easy, it might be difficult. We thought P2 is going to be showing uh dashboard, showing the uh interactions on the dashboard, showing news on the dashboard. uh those those kinds of items will fall in P2 bucket where P1 bucket is um we have a interface where we can train the uh agent on how to behave we'll have an interface where people will get live notifications if something is critical Satyasri Prabhakar Mantripragada: Okay. Yeshwanth Reddy Yerraguntla: so we want you to agree or disagree or argue with our logic Satyasri Prabhakar Mantripragada: If there is something that adds business value but is complex to implement then we should get started on it early so that we have time even if we fail. So if you marked it as P2 maybe if somebody picks it up at a later stage 01:56:37 Yeshwanth Reddy Yerraguntla: Okay. Satyasri Prabhakar Mantripragada: then we may not have time to reimplement if there's a failure or if we hit a roadblock. Yeshwanth Reddy Yerraguntla: I see what you mean. P we know it is going to be super complex to implement. Satyasri Prabhakar Mantripragada: I think Yeshwanth Reddy Yerraguntla: So we'll start in P 0 but we'll not deliver or we will not commit as a deliverable or not Satyasri Prabhakar Mantripragada: there should be some work happening on it. Vara Kumar Jagarapu: Yeah. Satyasri Prabhakar Mantripragada: Uh but you won't commit to the timeline on Yeshwanth Reddy Yerraguntla: to Satyasri Prabhakar Mantripragada: it. Vara Kumar Jagarapu: Yeah, let's say it's very important. Okay. P1 is also is there. Satyasri Prabhakar Mantripragada: Okay. Vara Kumar Jagarapu: Okay. P1. This might go a little slow also. Trade off P1 P2 when we have resources. Five resources. Okay. 51 let's say okay so P1 Yeshwanth Reddy Yerraguntla: I don't Satyasri Prabhakar Mantripragada: Okay, fine. Vara Kumar Jagarapu: that Yeah. 01:58:31 Satyasri Prabhakar Mantripragada: If there is something complex that takes maybe multiple iterations, Vara Kumar Jagarapu: Right. Yeshwanth Reddy Yerraguntla: Yeah. Satyasri Prabhakar Mantripragada: uh plan to have a maybe the squad lead or a senior member take over it. Um so that they'll also get the better picture and at at a point when Vara Kumar Jagarapu: Yes. Satyasri Prabhakar Mantripragada: they need to hand it over to somebody to actually implement it uh then uh it'll be faster. Vara Kumar Jagarapu: Yes, Yeshwanth Reddy Yerraguntla: But yeah. Vara Kumar Jagarapu: those are critical senior members. They'll be starting on that Yeshwanth Reddy Yerraguntla: Yeah. So P2 like for example if you take item five where Vara Kumar Jagarapu: P1 Yeshwanth Reddy Yerraguntla: news and insert life cycle that itself should be broken down into multiple tasks where some are known some are unknown. Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: unknowns P1 they have to be assigned to Manisha or senior. Vara Kumar Jagarapu: Mhm. Yes. And Yeshwanth Reddy Yerraguntla: Yeah. Satyasri Prabhakar Mantripragada: You got Yeshwanth Reddy Yerraguntla: So when we say P2 right now what we mean is P2 as a deliverable 01:59:34 Satyasri Prabhakar Mantripragada: it. Vara Kumar Jagarapu: there Yeshwanth Reddy Yerraguntla: but it it will in in a sense break down into tasks and those tasks will have P1, P2, P 0 as tasks. Vara Kumar Jagarapu: will be P2 only. But Start. Okay. Start Yeshwanth Reddy Yerraguntla: Um Vara Kumar Jagarapu: system. Yeshwanth Reddy Yerraguntla: outcome we just discussed that news life cycle generation as a P 0. Now there is no outcome in it but still we are calling it P 0 only. Right? Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: All I'm trying to say is P this this notion of P is all Vara Kumar Jagarapu: Okay. Cool. Yeshwanth Reddy Yerraguntla: about when are we going to start the work. Ideally speaking you should not touch any P1 until P 0 are done Vara Kumar Jagarapu: Cool. Add the Yeshwanth Reddy Yerraguntla: and you should not touch P2 until P one is done. Satyasri Prabhakar Mantripragada: I'm Yeshwanth Reddy Yerraguntla: But there will be tasks in P1 that are whose goal is P2 only. 02:00:45 Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: P2 tasks only. Vara Kumar Jagarapu: Clear. Clear. Satyasri Prabhakar Mantripragada: still Vara Kumar Jagarapu: Yeah. Fine. Fine. Yeah. Yes. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: Then also P2 but P1 Yeshwanth Reddy Yerraguntla: Correct. Correct. Vara Kumar Jagarapu: right Yeshwanth Reddy Yerraguntla: So yes. So any task which is PN will have subtasks uh from P 0 to Vara Kumar Jagarapu: yeah Yeshwanth Reddy Yerraguntla: PN minus one. Vara Kumar Jagarapu: agree. Yeah. Yeshwanth Reddy Yerraguntla: Good. Great. Vara Kumar Jagarapu: Okay then. Yeshwanth Reddy Yerraguntla: Um so right now we are still discussing only P in terms of deliverables only deliverables. Once we agree on them we'll then break them down into P1 P2 Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: whatever that is actually Vara Kumar Jagarapu: Uh then whatever you got assigned right that is Yeshwanth Reddy Yerraguntla: P2 that is P2 you can change it to Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: P2. Vara Kumar Jagarapu: Which one it is? 02:01:51 Yeshwanth Reddy Yerraguntla: Just kind uh 10. Row number 10. Vara Kumar Jagarapu: Okay. Yeah, this one. Yeah. Yeshwanth Reddy Yerraguntla: That is beautiful. Vara Kumar Jagarapu: Okay. Clear Yeshwanth Reddy Yerraguntla: Deliverable priority, task priority on coach. I mean if if we want to. Vara Kumar Jagarapu: and Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: and z12. Yeshwanth Reddy Yerraguntla: Great. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: So generate news life cycle broadcast news life cycle we are saying it's P2 Vara Kumar Jagarapu: Yeah, this is also P2. But yeah, Yeshwanth Reddy Yerraguntla: yeah Vara Kumar Jagarapu: P1US extract. Yeshwanth Reddy Yerraguntla: correct Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: correct alert we thought it will be Vara Kumar Jagarapu: Yeah. So Yeshwanth Reddy Yerraguntla: P1 only to implement high value. Yeah. Vara Kumar Jagarapu: already choosing same thing. Yeshwanth Reddy Yerraguntla: Ah so alert in UI alert in uh core agent. Vara Kumar Jagarapu: Yeah. Yeah. Yeshwanth Reddy Yerraguntla: So they will repeat. Vara Kumar Jagarapu: Cool. Yeshwanth Reddy Yerraguntla: Yeah. Same thing with training. 02:02:59 Vara Kumar Jagarapu: Then that is fine. Yeshwanth Reddy Yerraguntla: So training will we are saying P1. Vara Kumar Jagarapu: Nice. Okay. Yeshwanth Reddy Yerraguntla: Refresh is P2 or P3. Refresh is basically how we deliver news. Vara Kumar Jagarapu: refresh Yeshwanth Reddy Yerraguntla: So if news is P2, this has to be P. Vara Kumar Jagarapu: scheduleuler integration scheduleuler there should Yeshwanth Reddy Yerraguntla: Correct. Vara Kumar Jagarapu: be some chron jobs running that will just pass instructions to coet it's more independent Yeshwanth Reddy Yerraguntla: Correct. Vara Kumar Jagarapu: actually right Yeshwanth Reddy Yerraguntla: Yes. Uh, bro. Sure. Yeah. Yeah. We'll continue here. Vara Kumar Jagarapu: Yeah, this can but it is a valueologist. Yeshwanth Reddy Yerraguntla: I mean by itself it won't deliver any value this. Vara Kumar Jagarapu: It can it can be P2 only because like more reflect. Yeshwanth Reddy Yerraguntla: Oh no no Vara Kumar Jagarapu: So okay so it is again the P2 only then but this is an Yeshwanth Reddy Yerraguntla: correct. Vara Kumar Jagarapu: independent if someone is free maybe they can I mean junior developer can pick 02:04:25 Yeshwanth Reddy Yerraguntla: Yeah, Vara Kumar Jagarapu: it user preference let's Yeshwanth Reddy Yerraguntla: 100%. You Vara Kumar Jagarapu: say so it is also related to training Yeshwanth Reddy Yerraguntla: uh it is training kakapote. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: This is not explicit training in the sense that we are saying uh there's a trainer persona, right? every user uh the agent should adjust itself Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: and it's a very silent activity right so even if we even if our agent is not delivering uh user preferred format it is not a make or break in any way it's always a good to have so I will say this can go to P3 also Vara Kumar Jagarapu: Yeah, then that is fine. Yeshwanth Reddy Yerraguntla: governance is like uh like when I'm delivering news or when I'm delivering question answering uh I should never give uh certain answers to anybody like financial details. Vara Kumar Jagarapu: Yeah. Yeah. Yeshwanth Reddy Yerraguntla: So this is super crucial even for question answering this is super crucial even for P 0 this is very important but this by itself is it 02:05:48 Vara Kumar Jagarapu: H Yeshwanth Reddy Yerraguntla: needs a lot of research and other things. So um this is P 0 in the sense that what we will deliver might be very brute force or very inefficient in our way of organizing like certain users. um you will probably write a simple reex saying don't allow these kinds of patterns that in itself is not the right solution but to some extent it is doing right or we might say we'll have a Vara Kumar Jagarapu: Mhm. Yeshwanth Reddy Yerraguntla: hack by saying there is a final agent whose only like we are inventing a fifth agent whose only job is to ensure that uh that Vara Kumar Jagarapu: Check this all. Yeshwanth Reddy Yerraguntla: is again inefficient But in the interest of time and implementation, we might take that solution because uh it is brute force and but it is foolproof. It it will still work. It is P 0 in one sense but it is also P1 or P2 in a different sense. P P2 we can say by saying what we are delivering is a tech depth. 02:07:05 Yeshwanth Reddy Yerraguntla: We have to uh solve it Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: eventually. Vara Kumar Jagarapu: Uh not actually point of P 0, P1, P2. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: If there is a P3, Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: but when there is system already in place, system on top of it. So there should be something already existing. Yeshwanth Reddy Yerraguntla: H. Vara Kumar Jagarapu: Okay sir. Uh P1 P2 and dashboard system except this back. Uh Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: so testing key uh there should be some dashboard some data Yeshwanth Reddy Yerraguntla: H. Vara Kumar Jagarapu: running. So we can make it P2 or P3. But Yeshwanth Reddy Yerraguntla: Okay. My point is slightly different. During P 0 deliverable itself, people are going to ask questions which are going to be notorious. When I am saying I will deliver ad hoc question answering, Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: I cannot deliver it with uh without this feature. Vara Kumar Jagarapu: Okay. 02:09:11 Yeshwanth Reddy Yerraguntla: So we have to take some step whether it is efficient or inefficient. Vara Kumar Jagarapu: Okay. Deliverable clients Yeshwanth Reddy Yerraguntla: Client count. Vara Kumar Jagarapu: then then whatever you said is right. So at least like if you can get it one month of time let's say Yeshwanth Reddy Yerraguntla: H right that's what so this will Vara Kumar Jagarapu: but yeah Yeshwanth Reddy Yerraguntla: become P 0 plus P2 P 0 for one reason P2 for another reason so we have to put both of them on two I mean tap or maybe what you can do is I'll drop off in 10 minutes you can call Gopalar and tell this one exclusive Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: problem. Vara Kumar Jagarapu: Sure. Yeshwanth Reddy Yerraguntla: Knowledge sync life cycle is uh P2 because it is news related. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: Refresh insights and escalation uh is P1 because we are talking we are saying we'll deliver alerts. Vara Kumar Jagarapu: Refresh. Yeshwanth Reddy Yerraguntla: Refresh is P2. Vara Kumar Jagarapu: We kept it at P2. Yeshwanth Reddy Yerraguntla: I agree. 02:10:38 Yeshwanth Reddy Yerraguntla: But escalation is alert. So Vara Kumar Jagarapu: Escalation maybe. Yeah, as a subtask. Maybe we can keep Yeshwanth Reddy Yerraguntla: in yeah ah we can split it here. Vara Kumar Jagarapu: it. Yeah. and so only one thing we can check Yeshwanth Reddy Yerraguntla: Yeah. Uh, which one? Vara Kumar Jagarapu: this Yeshwanth Reddy Yerraguntla: Yeah. I'll have to drop off. Vara Kumar Jagarapu: share. No problem. Yeah. I I Yeshwanth Reddy Yerraguntla: So what you can do is create a fresh column Vara Kumar Jagarapu: will Yeshwanth Reddy Yerraguntla: saying P2 you'll have subtasks in P 0 P1 P2 that will dictate how to split the Vara Kumar Jagarapu: Yes. Yeshwanth Reddy Yerraguntla: tasks. Vara Kumar Jagarapu: Oh. Yeshwanth Reddy Yerraguntla: I think for every subtask you'll also have number of unknowns or how complex is the task itself that will decide whether juniors should take it or seniors. Vara Kumar Jagarapu: Yeah. Manisha, maybe uh we need your help over here. 02:12:08 Vara Kumar Jagarapu: Yeah. Thanks. Manisha Gundapuneedi: Bye Vara Kumar Jagarapu: Bye. Manisha Gundapuneedi: now. Vara Kumar Jagarapu: Yeah. Manisha. So the prior. Okay. Manisha Gundapuneedi: Oh, Vara Kumar Jagarapu: P2. There should be something that we need to do as part of present P1 subtasks. Manisha Gundapuneedi: no. No. Vara Kumar Jagarapu: then only we can deliver it into P2 right. So whatever you know right we can discuss on top of it. Manisha Gundapuneedi: Inc. Vara Kumar Jagarapu: Okay. Any Manisha Gundapuneedi: already. Vara Kumar Jagarapu: Okay. Manisha Gundapuneedi: I think Vara Kumar Jagarapu: Mhm. Okay. Manisha Gundapuneedi: research Vara Kumar Jagarapu: Thank you. Manisha Gundapuneedi: for Vara Kumar Jagarapu: Okay. Manisha Gundapuneedi: understand. Vara Kumar Jagarapu: Okay. But yeah, Manisha Gundapuneedi: for Vara Kumar Jagarapu: then maybe yeah. Ah, Manisha Gundapuneedi: Valentine. Vara Kumar Jagarapu: there Manisha Gundapuneedi: for Vara Kumar Jagarapu: Manisha P. Okay. Just I want to understand that is wrong. Okay. 02:16:04 Vara Kumar Jagarapu: So I thought like okay for Manisha Gundapuneedi: AB. Vara Kumar Jagarapu: example Manisha Gundapuneedi: So Vara Kumar Jagarapu: uh we uh Manisha Gundapuneedi: contagent Vara Kumar Jagarapu: Okay. So, Manisha Gundapuneedi: data source agent. Okay. Zero sticker key Vara Kumar Jagarapu: Right. governance life cycle Manisha Gundapuneedi: add expert. Vara Kumar Jagarapu: let's say multiple pri okay so Manisha Gundapuneedi: Um Vara Kumar Jagarapu: interface discussions on top of What needs to be Manisha Gundapuneedi: Um, Vara Kumar Jagarapu: implemented? So Manisha Gundapuneedi: for example, Vara Kumar Jagarapu: left. Okay. So, context Then Manisha Gundapuneedi: H Vara Kumar Jagarapu: that Manisha Gundapuneedi: context propagation is nothing but samehat. Start Vara Kumar Jagarapu: already start. Manisha Gundapuneedi: propagation. It is related to dat. So it involves multiple teams. dat. Vara Kumar Jagarapu: Manisha. Yeah. Yeah. Manisha Gundapuneedi: maint Vara Kumar Jagarapu: Silent sorry exploration. Okay. Manisha Gundapuneedi: Explore Vara Kumar Jagarapu: Yeah, not okay. Exploration problem. But yeah, um I got Manisha Gundapuneedi: before I put My Vara Kumar Jagarapu: multiple Manisha Gundapuneedi: thoughts in Vara Kumar Jagarapu: First of all, Manisha Gundapuneedi: basic Vara Kumar Jagarapu: right? Manisha Gundapuneedi: subtask. Unknown subtask. Vara Kumar Jagarapu: Yeah. announc. Right. Manisha Gundapuneedi: I already Vara Kumar Jagarapu: Right. Right. So okay. So clarity we need to achieve that clarity right. So, team Manisha Gundapuneedi: features. Vara Kumar Jagarapu: member Manisha Gundapuneedi: priorities. Vara Kumar Jagarapu: for example. Manisha Gundapuneedi: Oh Vara Kumar Jagarapu: Okay. Manisha Gundapuneedi: no. Vara Kumar Jagarapu: Whether we are able to give it to as a task to Manisha Gundapuneedi: Sorry. Okay. 132. Okay. Sorry. Okay. Bye. Transcription ended after 02:22:49 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.