Enterprise Brain Scrum - 2026/03/18 19:58 PDT - Notes by Gemini Mar 18, 2026 Enterprise Brain Scrum - Transcript 00:00:00   Yeshwanth Reddy Yerraguntla: Hi Manisha Gundapuneedi: Only my background Yeshwanth Reddy Yerraguntla: computer. Manisha Gundapuneedi: or cheap air closure. Yeshwanth Reddy Yerraguntla: But she thinks it Manisha Gundapuneedi: Play. Harry. Yeshwanth Reddy Yerraguntla: with Manisha Gundapuneedi: foreign. Yeshwanth Reddy Yerraguntla: 2028. Manisha Gundapuneedi: Okay. Vara Kumar Jagarapu: Tourist Yeshwanth Reddy Yerraguntla: six Manisha Gundapuneedi: Six Yeshwanth Reddy Yerraguntla: months. Vara Kumar Jagarapu: Say Manisha Gundapuneedi: months. Yeshwanth Reddy Yerraguntla: By the way, we are recording. Manisha Gundapuneedi: That's okay. Vara Kumar Jagarapu: out of me. Session ended after 00:02:51 00:13:10   Yeshwanth Reddy Yerraguntla: Right. Right. Vara Kumar Jagarapu: manager. Yeshwanth Reddy Yerraguntla: So you're asking uh these best practices. So UV for both uh package management and environment management and Vara Kumar Jagarapu: Uh okay. Yeshwanth Reddy Yerraguntla: rough ruf for linting. Vara Kumar Jagarapu: Uh-huh. Yeshwanth Reddy Yerraguntla: Adikura is from the same company which created UA. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: Um what anything else? Pi test for testing. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: Uh rough linting, formatting, Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: everything rough. Vara Kumar Jagarapu: Mhm. Clear. 00:13:52   Yeshwanth Reddy Yerraguntla: Uh setup is Vara Kumar Jagarapu: Okay. Okay. Yeshwanth Reddy Yerraguntla: called Rajashekar G: overall repo package. Vara Kumar Jagarapu: H. Rajashekar G: Okay. Okay. Vara Kumar Jagarapu: Initial stages. Yeshwanth Reddy Yerraguntla: Yeah, rough for linting and formatting. So, what I'll suggest is uh uh yeah, in the have uh make file. Rajashekar G: Are they Yeshwanth Reddy Yerraguntla: Okay. Make how make is just Linux specific not Vara Kumar Jagarapu: Python. Yeshwanth Reddy Yerraguntla: specific to any language. No, Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: I will ask a different question. How were linting uh and pretty pretty handled in JavaScript. Vara Kumar Jagarapu: configuration. Yeshwanth Reddy Yerraguntla: Okay. Vara Kumar Jagarapu: There will be CLA command before committing. Yeshwanth Reddy Yerraguntla: H. Vara Kumar Jagarapu: It will Yeshwanth Reddy Yerraguntla: So someone has to run it manually right that's what you're saying okay Rajashekar G: Nothing. Vara Kumar Jagarapu: pre Yeshwanth Reddy Yerraguntla: generally Vara Kumar Jagarapu: it will show there. Yeshwanth Reddy Yerraguntla: yeah say yeah right linting will show the linting issues and uh save format on save option if you select it will 00:15:42   Vara Kumar Jagarapu: Yeah. Yeah. Automatically it will. Yeshwanth Reddy Yerraguntla: automatically Yeah, Vara Kumar Jagarapu: Yes. Yes. Yeshwanth Reddy Yerraguntla: we can we can we can enforce it in the team. I don't know like how to enforce it at a or level or this kind of a thing so that people don't know Vara Kumar Jagarapu: Uh it should be at level. Uh we need to set up one time. Yeshwanth Reddy Yerraguntla: like Vara Kumar Jagarapu: I need to make Rajashekar G: Best practice engineering best practice. Vara Kumar Jagarapu: sure Yeshwanth Reddy Yerraguntla: Okay. Rajashekar G: Uh, Yeshwanth Reddy Yerraguntla: Okay. Okay. Vara Kumar Jagarapu: on top of it. came and Yeshwanth Reddy Yerraguntla: Yeah. My only concern is should we write a pre-commit Rajashekar G: heat. Vara Kumar Jagarapu: chooses. Yeshwanth Reddy Yerraguntla: uh uh code? Pre-commit is basically when when you're committing itself all the linting and formatting automatically Mhm. Right. Vara Kumar Jagarapu: sometimes. So that is not allowed 00:17:18   Yeshwanth Reddy Yerraguntla: Okay. Vara Kumar Jagarapu: for Yeshwanth Reddy Yerraguntla: I think rough log even in rough there is a similar configuration that once you create the tol file you write the rules for linting or pification Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: you allow these you don't allow these and you can do Anything Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: else? Vara Kumar Jagarapu: Yeah. Actually minimum Yeshwanth Reddy Yerraguntla: Oh yeah, we we missed that. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: Yeah, true. Vara Kumar Jagarapu: disc. Yeshwanth Reddy Yerraguntla: Important. In my opinion, those meetings are far more valuable than this share. Rajashekar G: Just checking Yeshwanth Reddy Yerraguntla: I can't see. Rajashekar G: current. Vara Kumar Jagarapu: Yes, maybe. Let me see that first. Yeshwanth Reddy Yerraguntla: Yeah. Vara Kumar Jagarapu: So, let's say emails related. Okay. So first of all at least for a project and it's a kind of system level. Rajashekar G: screen. Vara Kumar Jagarapu: Okay. Rajashekar G: Can you Yeshwanth Reddy Yerraguntla: computer Rajashekar G: share Vara Kumar Jagarapu: I'm sure 00:19:28   Yeshwanth Reddy Yerraguntla: EB plan. Rajashekar G: questions 1803 Yeshwanth Reddy Yerraguntla: Okay, I'm opening it. Rajashekar G: sheet you are in the yeah this Yeshwanth Reddy Yerraguntla: Got it. Rajashekar G: one so first four are like uh general seventh point the Vara Kumar Jagarapu: that Rajashekar G: technical DV related or how we need to Yeshwanth Reddy Yerraguntla: Okay. Sure. Which data source are we going to first fulfill in milestone one? Again, this is a business call. Currently, we have Gmail address. This uh if I have are you asking me to answer this or will is it okay if Raakasar answers this? Vara Kumar Jagarapu: blessing. Rajashekar G: Uh, someone has to answer. Yeshwanth Reddy Yerraguntla: Yeah, Gmail and uh Kaggle data sets is basically you're talking about manufacturing, Rajashekar G: Thank you. Yeshwanth Reddy Yerraguntla: right? Rajashekar G: Yes. Vara Kumar Jagarapu: Veracular children. Yeshwanth Reddy Yerraguntla: Gmail, Oracle or what else? Gmail. Rajashekar G: Gmail, G-Chat also can be there. Yeshwanth Reddy Yerraguntla: So the point of no we can definitely have Gmail G drive it's just that um if you handle one of them properly the other two should come for almost 00:21:16   Rajashekar G: All Yeshwanth Reddy Yerraguntla: free Vara Kumar Jagarapu: discussion there is some Rajashekar G: right. Yeshwanth Reddy Yerraguntla: correct Vara Kumar Jagarapu: CLA but it's not officially listed by Google but people are saying like we can use Yeshwanth Reddy Yerraguntla: Correct. Vara Kumar Jagarapu: it yeah we need to take a call but the feasible to when I'm Yeshwanth Reddy Yerraguntla: Correct. Vara Kumar Jagarapu: thinking is right So uh let's say a group of people are there right so when we are receiving email so individually we can receive it like but we should have all the credentials with us uh Yeshwanth Reddy Yerraguntla: for every one of them, right? Uh GWS, I think that was the only main Vara Kumar Jagarapu: um and whenever we are trying to Yeshwanth Reddy Yerraguntla: issue. Vara Kumar Jagarapu: save the arc snapshot right so how we are going to consolidate Those are Rajashekar G: So yesterday we thought like in a snapshot the details will be saved as project one summary Vara Kumar Jagarapu: emails. Yeshwanth Reddy Yerraguntla: Correct. Rajashekar G: project summary like that right Gmails are like person based if 10 users are there 00:22:22   Yeshwanth Reddy Yerraguntla: Correct. Rajashekar G: 10 users Gmails are syncing with me so how are you doing Yeshwanth Reddy Yerraguntla: Thinking in the snapshot. Rajashekar G: no initial at syncing Gmail means so there are 10 users 10 users Gmails will be synced right. Yeshwanth Reddy Yerraguntla: Cut it. Rajashekar G: So how the system will know like uh which which emails are from which project or how we are going to link Yeshwanth Reddy Yerraguntla: I mean that will that information you'll get while reading the Rajashekar G: it. Yeshwanth Reddy Yerraguntla: email right when you do a hashtag you know which email belongs to which project and which persons it it belongs to so during the sync job only it will generate that summary saying this last five minutes these 100 emails came this is the entire content email one this is the content email two this is the Or it might uh say email 1 2 3 4 5. This is the summary. Vara Kumar Jagarapu: I put it life cycle. Uh let's say okay so data Yeshwanth Reddy Yerraguntla: All Vara Kumar Jagarapu: sources with some credentials actual 00:23:56   Yeshwanth Reddy Yerraguntla: right. Vara Kumar Jagarapu: data first of On top of it, Yeshwanth Reddy Yerraguntla: H Vara Kumar Jagarapu: it should deterate and specification. What is waterp? Yeshwanth Reddy Yerraguntla: Yeah. So every person let's say there are 10 people in the organization. Every person got five five emails. Yeah. is now I'll show my own email box right so few emails are leave related few emails are uh uh whatever these are promotional materials some are related some are daily digests so in the 10 events there is a very good chance that eight of them are useless Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: So yeah then for me but for Ashwant these are the important emails. Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: This one this one uh this probably we can ignore because we have a Jira but we can have this also this one. Mostly for me it my important emails are recordings. Sometimes there will be client meeting related to these things Vara Kumar Jagarapu: H. 00:25:44   Yeshwanth Reddy Yerraguntla: genrx. So uh you already have that categorization here itself. So enterprise brain I have two different uh mails. Uh so you'll have a text file saying uh uh enterprise uh brain summary. So, email one had this, this, this, email two had this, this this, this, or it might be like a more naturally flowing summary. Um, so and so happened at 8:30 p.m. Then it might do some linking saying first email um I mean this information is coming from email one some other uh uh thing happened at 9:30 that is the second email based on that uh uh this is the overall Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: summary. So this is coming from exclusively H1's email for enterprise brain. Vara Kumar Jagarapu: Mhm. Yeshwanth Reddy Yerraguntla: So H1's email for gen HRX will be a similar uh thing. Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: Correct? Now this is only for S1. Now this way you will have 10 more people. 00:27:14   Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: Correct? Yeah. The the point of data source agent is to give this much and let the real time agent consume this and then decide how to group it. If enterprise brain itself has different uh males from different Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: people uh uh same content this now it is trying to summarize uh for enterprise brain now it is trying to group by enterprise brain so happened at 8:30 according to H1's email uh according to XY Z email these are more items Right. So now this is generated by uh RT agent at a of this level. Vara Kumar Jagarapu: Good. Yeshwanth Reddy Yerraguntla: RT agent will anyway have snapshot of yesterday's enterprise brain. Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: So it can yesterday's uh this thing only snapshot only because we are syncing every 5 minutes it might fetch today's snapshot um uh Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: update the today's snapshot with this information you have if I take this paste it here let's say this is today's snapshot at uh uh 8:00 a.m. the same. 00:28:54   Yeshwanth Reddy Yerraguntla: Now you got a new snapshot at 8:05 a.m. For an LLM, Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: it's not hard to read this text file, this text file, and create a third text file. Vara Kumar Jagarapu: H. Yeshwanth Reddy Yerraguntla: That's what it will do. That is how you constantly update the snapshot. Uh, am I not making sense? Rajashekar G: I know one doubt here like if there are uh same point we have written in the sheet also if there are like uh hundreds of users there will be that huge data let's say Yeshwanth Reddy Yerraguntla: Got Rajashekar G: 10 like each person let's say he has received 50 emails like Yeshwanth Reddy Yerraguntla: it. Rajashekar G: 500 if there are 100 or 200 users that will blow Yeshwanth Reddy Yerraguntla: Got it. Rajashekar G: color. So how that going to Yeshwanth Reddy Yerraguntla: So you will have to create temporary Rajashekar G: handle Yeshwanth Reddy Yerraguntla: intermediate uh summaries of emails with respect to some grouping criteria. So one criteria could be the projects. Again you constantly think in the same dimension. 00:30:19   Yeshwanth Reddy Yerraguntla: You have projects, you have pipelines. So even if you have thousand emails to sync, if you can write these thousand emails in some temporary file storage at a folder level saying enterprise brain, these are the 300 emails I received. uh for something else. These are the uh uh 50 emails I received. If you temporarily store them to process them on a folder by folder basis will be more easy for it to handle. We have to do it again. Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: Definitely we are dealing with a scale issue here and uh for a project like enterprise brain where the entire organization needs to be synch right I'm able to come up with some strategy because I am constantly asking myself the question if I were to handle this myself if I am at AI how what is the best way I would want to consume this information. So constantly think in terms of temporary files. Rajashekar G: Right. Yeshwanth Reddy Yerraguntla: Now let me first uh read all the emails. 00:31:44   Yeshwanth Reddy Yerraguntla: Too many emails reading at once is overwhelming for human being also. So I will first store all the emails in buckets. Let me do only that first. Vara Kumar Jagarapu: what? Yeshwanth Reddy Yerraguntla: Then in each bucket I'll go read all the emails. My uh head space is very clear. I can either use single agent to read all buckets one after the other or I can use five agents to read five buckets parallelly. Why parallelly? Because every bucket we are guaranteeing that the information will be independent of the other. Rajashekar G: All Yeshwanth Reddy Yerraguntla: That is how we have to think rancher. Rajashekar G: right. Yeshwanth Reddy Yerraguntla: We don't have a choice. Vara Kumar Jagarapu: So yeah. Yeshwanth Reddy Yerraguntla: For the time being, yeah, we can simply need not worry about that scale Rajashekar G: Okay. Yeshwanth Reddy Yerraguntla: or Rajashekar G: So it's like syncing means right now what we need to do if there are 10 Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: H. Rajashekar G: users 10 users emails on that particular uh day will be dumped into the system and based on that RT agent will analyze and break them 00:32:54   Yeshwanth Reddy Yerraguntla: Got it. Rajashekar G: into project level categorizations then update the snapshot Vara Kumar Jagarapu: Come on. Yeshwanth Reddy Yerraguntla: Yes. Rajashekar G: Are we missing anything Vara Kumar Jagarapu: You know, just Rajashekar G: here? Vara Kumar Jagarapu: specific. It is only organation specific and user specific storage always. Rajashekar G: We need to apply the back that needs to be tagged with some ID user ID without link Vara Kumar Jagarapu: Yeah. Rajashekar G: over. Vara Kumar Jagarapu: Specific summary, right? Yeshwanth Reddy Yerraguntla: Right. Rajashekar G: Oh no. Yeshwanth Reddy Yerraguntla: Give me a check. Vara Kumar Jagarapu: personific. Yeshwanth Reddy Yerraguntla: Right. Vara Kumar Jagarapu: Okay. So next for example Yeshwanth Reddy Yerraguntla: How cool. Vara Kumar Jagarapu: So, snapshot and Yeshwanth Reddy Yerraguntla: If it's a whole text file, Vara Kumar Jagarapu: uh Yeshwanth Reddy Yerraguntla: you're saying it's all mixed up Rajashekar G: H. Yeshwanth Reddy Yerraguntla: now. Rajashekar G: Yes sir. If it is a textual mix project enterprise project having all the males of Yeshuant, Prabhakar, Naven, me uh even junior developers males also will be Vara Kumar Jagarapu: Okay. Rajashekar G: there. 00:35:08   Yeshwanth Reddy Yerraguntla: All Rajashekar G: So we need to have another chunk like let's say one column is project, Yeshwanth Reddy Yerraguntla: right. Rajashekar G: one column is user, then the summaries will be again Vara Kumar Jagarapu: Privacy. way of thinking maybe at least okay at least project information that can be shared. Rajashekar G: All Vara Kumar Jagarapu: So there are some personal mails also I mean but yeah Rajashekar G: right. Vara Kumar Jagarapu: let's say if I'm reporting something to someone feedback so Yeshwanth Reddy Yerraguntla: Sorry just for this one problem. Vara Kumar Jagarapu: Just thought them Yeshwanth Reddy Yerraguntla: So again my head is going in a direction of saying okay text file if it is too difficult to handle we can say we'll start dealing every uh point as a bullet point end of the day everything is just a bullet point in this For the bullet points, you'll have uh these hashtags again. These are the use um these people received the email or these people were Rajashekar G: You're Yeshwanth Reddy Yerraguntla: involved like they were involved Rajashekar G: right. Vara Kumar Jagarapu: H. 00:37:09   Yeshwanth Reddy Yerraguntla: in that email. Uh next hashtag can be talking about the qualitative thing. This is coming from uh Jira. This is coming from uh #acha dollar. I'm just giving some notation source dollar hashtag will be Rajashekar G: Let me Yeshwanth Reddy Yerraguntla: qualitatively saying this is a here it could be # Rajashekar G: see. Yeshwanth Reddy Yerraguntla: finance # security deadline delivery. We will decide these hashtags. Vara Kumar Jagarapu: What's Yeshwanth Reddy Yerraguntla: What then happens is uh when you fetch the uh items for let's say now you want to deliver something to Navin uh naturally Vara Kumar Jagarapu: that? Yeshwanth Reddy Yerraguntla: Naven will be interested in hashtags with delivery uh hashtag with engineering engineering uh like these if we can sort of treat these as verticals these are sources these are people If we can group them like that, Vara Kumar Jagarapu: Amen. Yeshwanth Reddy Yerraguntla: it becomes a structured data problem. Then we putting hierarchy, putting security will become trivial. Uh just a thought. Rajashekar G: So tags will be in the same text format when you want it. 00:38:44   Yeshwanth Reddy Yerraguntla: I'm not saying we'll do it that way. This can be a JSON JSON of list of items or this this itself is a JSON where you have uh content is this key. Uh senders and receivers is this key. Source is this key. Um content type is this uh category is that Rajashekar G: Getting Vara Kumar Jagarapu: Okay. Yeshwanth Reddy Yerraguntla: key Vara Kumar Jagarapu: Simple question. Let's say uh client is there. Okay. For example, Yeshwanth Reddy Yerraguntla: H are you asking me? Huh? He can read because in somewhere somewhere else hierarchy we will say uh for the name Naven these are all the possible people Naven can already read. Vara Kumar Jagarapu: Um roles for Yeshwanth Reddy Yerraguntla: Uh oh no. Oh Vara Kumar Jagarapu: So it is Yeshwanth Reddy Yerraguntla: this instruction Ninsar gave us clearly that they should be able to Vara Kumar Jagarapu: something. Yeshwanth Reddy Yerraguntla: read. Vara Kumar Jagarapu: Okay. So then Yeshwanth Reddy Yerraguntla: We can toggle it. 00:40:17   Vara Kumar Jagarapu: okay. Yeshwanth Reddy Yerraguntla: As a company head, I should have the facility to read anybody's email. Vara Kumar Jagarapu: Okay. Another Yeshwanth Reddy Yerraguntla: Gmail. Vara Kumar Jagarapu: question, Yeshwanth Reddy Yerraguntla: Yeah. Privacy problems. Vara Kumar Jagarapu: privacy or compliances, whatever it may be. Yeshwanth Reddy Yerraguntla: Are they Vara Kumar Jagarapu: Chases. This Yeshwanth Reddy Yerraguntla: if the toggle is on then we can maintain that hierarchy saying whatever arbback saying nav as a role can access um everything that is accessible by this principle and whoever is below below uh Gopal Anund Gopal itself will have its own rback in the sense that Gopal can read Wenut eshwant has an arback of uh Rajar and Manisha. So if you recursively go Naven can read Gopal, hence Navin can read Wankut and Ashwant. Hence Naven can read Raja and Manisha. Vara Kumar Jagarapu: Yeah. Yeshwanth Reddy Yerraguntla: Sorry, Vara Kumar Jagarapu: For example, Yeshwanth Reddy Yerraguntla: Huh? with the toggle. You can do Vara Kumar Jagarapu: with a toggle but yeah we are giving a scope because let's 00:42:16   Yeshwanth Reddy Yerraguntla: it. Vara Kumar Jagarapu: say Whatever the Gmail built the privacy Yeshwanth Reddy Yerraguntla: I agree. But some manufacturing companies, some nonIT companies may need it. Some companies may not need it and in the case where you don't need it, we are not even giving that feature. This is where you will stop it for Gmail for security sensitive sources. We will not open it up for nonsecurity sensitive sources like Salesforce or Jira. So Jira let's say is not at all security sensitive then you can open it up for everybody. Vara Kumar Jagarapu: Okay. Okay. Yeshwanth Reddy Yerraguntla: As long as this is dollar jer this will apply. As long as it is dollar Gmail this will not apply. That also we can do and that is in our hands whether you want to at a source level do you want to keep the arbback sensitivity or not. Rajashekar G: Now we have to join generics Vara Kumar Jagarapu: Okay, Rajashekar G: call. Yeshwanth Reddy Yerraguntla: Okay, 00:43:40   Vara Kumar Jagarapu: we'll continue that and Rajashekar G: Yeah. Yeshwanth Reddy Yerraguntla: fine. Vara Kumar Jagarapu: yeah toggle. Okay. toggles something. Rajashekar G: And previously also an this point came like when we were building Gmail uh character in enterprise plan. Vara Kumar Jagarapu: Okay. Rajashekar G: So the entire point was the emails I involved only I should be able to read also like if I and Yeshwanth Reddy Yerraguntla: Another Vara Kumar Jagarapu: For example, Rajashekar G: you access someone shouldn't come and look into this part. Vara Kumar Jagarapu: When it comes to Ghat that Yeshwanth Reddy Yerraguntla: two minutes. Vara Kumar Jagarapu: information Yeshwanth Reddy Yerraguntla: Just 2 Vara Kumar Jagarapu: okay Yeshwanth Reddy Yerraguntla: minutes. Vara Kumar Jagarapu: So how these are going to be structured again. Yeshwanth Reddy Yerraguntla: We can do something like this. This at least this much we can agree upon, right? If we can break down every item into its own summary, lots of items we will discard because they are spam or not necessary. But those few items which are important, these are there we are calling them as bullets or some items. As long as we can maintain that uh we can trace from where it is coming we can always put the R back on them. So ethical should we expose it or not expose it? Let the business take the call. Rajashekar G: That call is ended. We can discuss again. Yeshwanth Reddy Yerraguntla: We can ask anyway we'll jump. Rajashekar G: Yeah. Transcription ended after 00:46:06 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.