Vexa: Your Firm's Most Valuable Data Is Trapped in Meetings

Dmitry Grankin (Vexa) (00:00)
meetings is something you cannot automate.

If we still meet and we decide and we build trust and we steer, that's the way to go.

Michael McLaughlin (00:26)
Welcome Dmitry Great to have you here on the Stack to Street podcast. when I started doing some research for this podcast.

where I attempt to make a connection between developers and the sort of front office executives that I have in my network as part of the finance community. I came upon Vexa and thought, now this is something pretty unique because the one thing that AI can't really take over is the organic meetings that take place.

in banking and finance and relationship building and deal doing and everything like that. So give me a sense for the problem you're trying to solve.

Dmitry Grankin (Vexa) (01:10)
companies run on meetings, right? And they will continue so. AI we'll not change that because meetings is where people build trust, right?

it's been the case like thousands years ago and it will continue so.

So yeah, that's exactly the thesis here, building knowledge from meetings, because meetings is the first tier communication that we have.

Michael McLaughlin (01:45)
So

imagine if you were talking to

Senior managing director, the head of capital markets at Morgan Stanley. what are the use cases that you would like to tell these non-tech leaders about?

meeting intelligence

Dmitry Grankin (Vexa) (01:59)
every communication starts with with the meeting. And at the same time it happens to be that those conversations are lost if uncaptured. And this is arguably most valuable

pieces of the piece of data. We have emails, we have documents, we have you know execution logs, transactions, it's all persistent and it all has its own place. Meetings meeting data doesn't have its own place. It's standard home. So that will be the standard place where your conversations get

stored, secure, air gapped, self-hosted inside your organization. It's the first thing. And it's delivered where you need it most, real time, any time. That that's the thing. The second thing is how to leverage that. We know it's valuable, we know that we everybody

now understands they they need meeting data. But what's what's the next thing? usually companies r rationalize

implemented implementing meeting bots is that they got a transcript and they feed it into LLM and get transcript and they store this transcript somewhere. But the problem is that all the details are lost.

if you have like twenty people on a meeting, like you have a CEO, CTO and other guys, and each of them will look differently at what's happening and what's going on. So you need at least you know you have five people on a meeting, you need at least five summaries.

So you need to own the the raw data so that you can interpret it all the time. But it better is that each of these individuals they built their own knowledge out of it. But all right, you build your own knowledge, you have like each of them has got their own repository.

It's also impractical because there is a shared layer. It's always a bunch of shared layers. Like there is like a global layer of of knowledge that you want like any agent, every in every person to know about the company they work for. there is a a group knowledge, like we have a recurrent meeting every every week, and we want

This knowledge to be built up and be shared. And like every organization's got natural boundaries for how their information is available.

Michael McLaughlin (05:11)
when I was in banking thinking about all the meetings, all of the Zoom calls, all the Google Meets calls, the Teams calls,

you know, you have a a one hour intense call. And then at the end of it, sure, there might be a transcription that's provided, but you know, what do you do with that transcription? what happens next?

Dmitry Grankin (Vexa) (05:32)
for a lot of products right now what they do is you have the tr transcript and you do a single pass into a LLM and you have a summary. But the problem is that it's isolated. So the LLM knows nothing about you.

it knows only the transcript and that's it. what we do instead we build knowledge with that. like we build it the like Wikipedia of company knowledge

Michael McLaughlin (06:08)
Vexa's foundation is open source, Tell me about that journey.

Dmitry Grankin (Vexa) (06:12)
And yeah, how it started, I was building a proprietary meeting note-taker, care. And being a developer myself, I found problem with understanding of what users actually want because if you build a proprietary product, your connection with the users is very loose. You deliver a product and they even use it or they don't use it.

With the open source, you have a developer and developer brings you a very high energy signal because someone has already paid this developer to find the solution. And so this developer comes to you and they run your tool, they find issues, they file those issues on GitHub.

and they might even contribute their code back to your code base. for anyone who's ever started an IT startup, knows how important is that to have a good bandwidth with your users, and that's the solution here.

Michael McLaughlin (07:25)
this meeting data the and this knowledge graph

that Vexa helps create can be ingested by an individual's workflow. let's face it, individuals

have their own workflows. I mean when when you sit at your screens at a at a bank, everybody's screens are different. You know, I might have had Microsoft Outlook in front of me because I'm an email guy, or I might have might have PowerPoint in front of me because I'm a presentation guy, tell me about how your infrastructure allows for

infinite amount of use cases based on the workflows of the individual.

Dmitry Grankin (Vexa) (08:05)
Yeah, that's nature of open source. You create infrastructure which is reusable your reusable components. if you need Outlook, you just build the Outlook extension, j just the Outlook integration for it. If you need Google Docs you build Google Docs

That's how open source works is that you don't wait for your vendor to deliver the thing. You can build it yourself.

Michael McLaughlin (08:32)
Yeah, I think that's I think that's really powerful.

so yeah

Dmitry Grankin (Vexa) (08:35)
Yeah.

Michael McLaughlin (08:35)
I wanted to then maybe

pivot to your commercial endeavor with Vexa there's an opportunity for you as a developer to to monetize this. And I wanted to hear about how you plan to do that.

Dmitry Grankin (Vexa) (08:49)
open source is quite commercial. Like it's a very established business model. even if a project doesn't belong to anyone, there's still a lot of vendors that are maintaining it making money out.

as a bank what you you care about is that you own the software so it works inside your stack you don't rent software you don't put your

Precious data outside.

And the second thing is that this brings a conflict. Is that you want to own it, but you don't want to babysit it.

the alternative solution is just to pay for someone who can who can deliver software with the right attestations with the right feature set while being at the same time fully self-hosted. that's one thing. The other thing is that companies want to have a stake in something they

depend on because open source becomes a dependency. They want to steer the process, they want to keep a vendor close. They want to have a bandwidth with vendor. Even if they contribute, they want the vendor to be able to to merge their

their features into upstream. You can boil this down into these things in or even into one thing is that if you this is your asset, your own asset, this is your dependency, you want to invest into it.

There's a bunch of industry foundations that make it possible. There is a Linux Foundation, and there are multiple branches of Linux Foundation for every industry that makes this collaboration between competitors essentially, companies that compete with each other to collaborate on the things that they share.

as infrastructure.

Michael McLaughlin (11:05)
If managing director at a at an investment bank is listening to this podcast and they're saying, I don't know, man. I use Zoom, it transcribes when I think about it, I save the transcription down into my file folder. And you know, maybe I go back and look at it. Maybe I upload it into Salesforce, you know, those sorts of things. So I'm I'm all good.

Dmitry Grankin (Vexa) (11:26)
we have a few meeting platforms like Google Meet, Microsoft Teams, Zoom, and we also have meeting rooms and each of these environments has its own vendor, its own platform and

They are not compatible with each other. Like that's why you have to maybe choose between Zoom or Microsoft Teams.

Michael McLaughlin (11:50)
Yeah, Zoom doesn't

talk to Meets and Meets doesn't talk to teams.

Dmitry Grankin (Vexa) (11:54)
No, no, doesn't doesn't talk. So you have to choose between whether you stick with Microsoft or you stick with Google. And you get into their stack, into their environment, you become dependent on a proprietary vendor,

Michael McLaughlin (12:14)
there is something to be said for

keeping your data in your GPUs, right?

Dmitry Grankin (Vexa) (12:19)
yeah it open source give you gives you that you can even have a full air gap

meeting experience you have open source meeting platform supported by Vexa so you can it's have fully air gap meeting platform inside your stack.

Michael McLaughlin (12:40)
when you're an investment banker and you're doing financing for someone, they're telling you everything. for a bot to sort of show up in the dialogue from some company is a little strange, right?

Now if you can sort of say to your clients, Yeah, we've got the bot that shows up, but it's our bot. it's just like you're talking to me. This is just my assistant and it won't go anywhere.

Dmitry Grankin (Vexa) (13:04)
Yeah, yeah. so you know it's yours and there is an LLM part to it, which

you can self-host as well. Like we have we have a customer which is a bank, a European bank, and they self-host a bunch of LLMs

And yeah, you can be sure that whatever happens, it's all there. you can eventually just go and delete it. If if it doesn't doesn't go the right way. yeah.

So you're clean here.

Michael McLaughlin (13:39)
it would seem to me that Vexa is actually being developed in an environment that's sort of heading your direction, the direction of on prem, the direction of self hosting, the direction of of having your own LLM, if you will.

Dmitry Grankin (Vexa) (13:57)
and banks all have this open source model now deployed. So they have

models they have they all have transcription models. Like they own Whisper. They all do. They all have open source Chinese models deployed.

And they all have Kubernetes and all the basic infrastructure Vexa is, you know, just a glue layer, just it's just a layer on top of of this stuff. every piece of software is a combination of open source. Vexa is all is a combination of underlying infrastructure as well.

we are now getting on boarded to FINOS which is a part of Linux Foundation in f for financial industry. and there is a bunch of organizations there which which will get involved after we are properly on boarded there.

So it's a it's a publicly available information but not yet announced properly. So the the way to to get engaged and to contribute is just to join one of the FINOS meetings that we'll that we'll have.

There is a kind of this S-curve of adoption that's just getting started right now. Everybody is building its own thing in their garage. in five years we end up with industry standards, with a few industry standards. We want to be one of these industry standards.

for enterprise knowledge management, enterprise intelligence, and it's all getting started with meetings because meetings is something you cannot automate. You cannot automate for a simple reason. We cannot automate human responsibility, human authority.

If we do that, we are lost we lost the battle against AGI. If we still meet and we decide and we build trust and we steer, that's the way to go.

Michael McLaughlin (16:14)
Yeah, that's that's really, really well said, Dmitry and I agree.

Before we we break, I wanted to I wanted to learn about how people can be in touch with you

Dmitry Grankin (Vexa) (16:28)
there is a GitHub repository. The natural way is that you just find the repository exactly as as you found me. you popped up on Discord and you wrote me an email and th that's how it's getting started or you just find my website which is Vexa.ai

dot ai and click talk to founder I'm happy to talk to domain experts about enterprise intelligence, meeting intelligence.

Yeah, that's

what we do all the time.

Michael McLaughlin (17:07)
I'll put those links and that information in the in the show notes and in the description of the podcast so that when we launch the podcast on all the distribution channels, someone can can click and be in touch with you.

Dmitry Grankin (Vexa) (17:20)
it would be awesome. It would be awesome, actually.

Michael McLaughlin (17:23)
All right, my friend. I really appreciate it. Thank you so much for your time. And let's see if we can we can get some some some high quality stars from this.

Dmitry Grankin (Vexa) (17:32)
Awesome Michael. That's great conversation and I appreciate I appreciate your interest and your time. Spent thanks for having me.