Amit Eyal Govrin, Kubiya.ai | KubeCon + CloudNativeCon North America 2023

This is Textron tv. Hey there everybody. Welcome back to Chicago to Kku coupon 2023.

tv, and, uh, cloud native now. So we've got more great interviews. Alan's just finished a couple and I'm teed up now with Amit girlfriend.

Nice to meet you. Very nice to meet you. And you are with, say the name Kbi ai.

Kbi. ai A ai, okay. That's, that's a lot of syllables there, but, okay.

I'm just Kidding. We do this on purpose to confuse, You know what, once, once you hear it, it's unique. Right?

You remember those a hundred percent. Exactly. Okay.

So I know you've talked with this before and we've talked about, you know, IDPs and, you know, it's, it's, it's a really kind of fast growing area, right? Everybody's focused on Absolutely. How do we help developers, platform engineers, the people all working together to mm-Hmm.

Kinda make lives easier, self service, that kind of thing, and portals and all of that. Uh, I know you, you had some announcements, uh, recently talk about that, but maybe kind of set it up of, of what you all do and why that's a little different. Sure.

I, I think maybe to take one step back, uh, one of the things that, uh, I think a lot of, uh, companies in the industry, including ourselves, are doing is how do you create the, a better exchange between the developers and the DevOps, the operators on one side, developers on another hand, where you have the golden path towards productivity. Uh, IDPs, internal developer platforms have been known, uh, backstage being kinda the primary leader in this space, uh, to unlock lab productivity. One of the things that we at Kuia have discovered, and this is what we've tried to essentially address, is the self-service is only as good as the interaction between the users and the tools.

And where often it does break down is on the fact that it's a one directional exchange between the user and the tool. If you fail on authentication, fail permission, don't know the variables you need to pass or anything, just doesn't work as expected, you have a question. You have to go back into Jira ticket queue in order to go to essentially get assistance from a human being.

So what has come to, to fruition in the last year, obviously Chad, GPT doing the market education for this is conversational AI having a two directional exchange between machines and humans. We've married these two worlds together. We've created that exchange between your users, developers, and operators with the tools in a way where you can have a full on conversation with it.

You can go and spin up a Kubernetes deployment, it will ask you how many you replicate. You can scale it up, scale it down. You can have, uh, where your Jira ticket queue, for example, and ask you questions about who owns what, where, where they're assigned creation, notating, anything that you would be able to do as a human only in this case, you're just having a conversation because imagine having on the other side, someone like Fred, John, Bob on the on-call channel, having a conversation with you.

We're replacing now with the virtual assistant. Interesting. So, um, so replacing is the virtual assistant model.

Yeah. It it sounds like also you were maybe describing if I got this right, that the developer, uh, technical staff can interface through chat GPT to say, I mean, nice thing about chat GPT is you can talk to it like a programming language just in, you know, conversational logic, right? I want you to create this for me with blah, blah blah.

Can you, is that part of what you're saying? You can kind of give it instructions or requests One minor re ation, we're not using tragedy. Oh, Okay.

Okay. It's a tragedy like experience. See, it's probably the best way to anchor people to, like I jumped, I jumped that, Which is fine.

We're using about eight different language models, both large and well. We're actually using conversational AI piece with um, uh, GPT four for Azure, just kind of AF yi, but that's separate. The experience is Chad, GPT, like only here.

It's your tools, your programs, your scripts, uh, essentially a personalized IDP if you may, for each and every persona in the organization. Each and every individual in the organization. Okay.

So, uh, are you using generative AI large language models underneath or that's just the experience that you're delivering with whatever technology? Certainly. So large language models are giving that natural exchange, uh, essentially the benefits of large language models and the generative piece we're using, but we're also harnessing and scaffolding it with rule-based systems with sound engineering, uh, practices.

So that's how you get the benefit of an expected outcome. On one hand, uh, our users wouldn't, would expect an expected outcome essentially has to be deterministic. What they'd get on the other end That, I'm glad you said that 'cause I was getting kind of thinking down the chat g PT line.

Sure. It's nice, but you don't get the same response to, you know, same outcome, not necessarily ga guaranteed to be consistent, to be accurate. So it makes sense on the backend scaffolding Department and there's permissioning and rule-based systems in place.

Yeah. Uh, to harness the LLM to, to function the way you would expect it to. Very nice.

So you talked about the developer experience and, and I totally get the cognitive distance dissonance between I'm working in my, you know, IDE and doing my developer thing. Mm-Hmm. And then I gotta go to this ticketing system or go to some weird kind of operating portal that doesn't, wasn't built for developers, but I still gotta use it.

Maybe It is, but it's, but it's only as good as how it's been upkept maintained or what you've actually input into there. So you still have to go to program it to act the way you expect it to. Well, and it's, it's part of the flow too, right?

Yeah. I mean, I might not know all 10 answers to all 10 questions at the start of the interaction. Yeah.

But I need to get it started and then Mm-Hmm. You know, finish it at some point in the process. So, but it's about in the flow of the developer work, not I'm a form filling me out.

Right. Uh, a hundred percent. Great.

Well, so, so the announcement then, summarize again for us, 'cause I kind of stepped on it with the GP Two stuff. Yeah. Uh, no, it's, it's fine.

So we, we came out last year actually with our initial, uh, platform, which was more resembling, um, next generation of workflow automation because it was more so to get people connected to the world of conversational AI into the world of, uh, IDPs. Right? Where we took it from there, and especially with evolution of large language models, is an agent based approach where we're using different models and different agents trained on different models in order to map out effectively different functions within a transaction, almost like a human brain in the lobes interacting with each other.

So when you go and interact with the system, depending on how you approach it, it could be show me, show me how I do something. Say set up AVPN that may be in your docs. So we are using rag techniques obviously to go into a train on your, on your knowledge base.

But if you're gonna go and run a query against a data source or database, or maybe show me all my Jira tickets for current Sprint for certain, that's a different language model that's interacting with you. Mm-Hmm. And then if you're gonna go and create a Terraform, create a Kubernetes deployment, destroy Terraform, that's an action base that's also needing to go into verify permissions.

We are going, we're using open policy agent, uh, we're also using all the techniques to data masking, uh, and then we obviously have the orchestration. They're a state machine that will go into orchestrate all of those with the business logic of the organization. So there's a lot going on behind the scenes.

It is, well, a lot of different kinds of routine to the user request. Yeah. It's seamless.

You, you have a brain emoji thinking and then reading back to you what it's doing. So it's actually very fun interaction. Human-like interaction, but then behind the scenes there's a lot that's going on.

Very cool. Um, do you integrate at all with IDs and kind of go that far with it? Or is it a kind of separate users Can integrate with their IDs?

They can go and plug in, we have ASDK and it's all open API based. So essentially you can connect to any tool That's, I think it's really nice and important because well, one, it's, it's right there handy, you know? Yes.

And also if you're delivering any information that you can kind of deliver in the way that code looks like or structured or autofill or you can kinda mimic the behaviors of what you do in an IDE that Mm-Hmm. Very natural. So I don't have to go look up an API for what that call is.

It's, it's right there. Or I don't have to go look up what that thing's called on my request. It's right there.

Agreed. So I think maybe the one call out I, I would love to make is people are thinking about the co co-pilot approach, or Chad, the approach as a code completion tool. Mm-Hmm.

What we're doing is we're doing an operations, uh, completion, essentially all of the next step. What would you do? How would you deploy the code, run your CIS CDs right?

Um, and maintain it for day two operations. So that's kind of where you can think about us filling that gap. Okay.

Interesting. Well, so congratulations on the announcement. That's fantastic.

And you know, every time you come, you've got new and progress and things happening. I, I'm curious, maybe you've talked to Alan about this, or Mike, sort of your backstory. You're CEO founder of the company, right?

How did you get going down this path? Was that I was doing this at a big company and I rec No, no, this is like a huge need. I'm gonna go Forward.

It it's a great, it's a, it's a great question. So the first time we we came on, it was myself and my co-founder, who's he himself is an operator who built a self-service platform for his platform team in his previous company at BlueVine. And I came from AWS where I was managing partnerships for all the DevOps and SecOps.

Nice, uh, tier one players. So I, I saw where they were possibly struggling with some of the self-service and where I was in the middle of trying to assist them. And that's kinda where the thesis on the business side, which is a, this is a much larger problem and how we've actually solved this ourselves kind of came together.

Interesting. Talk a little bit about the, your experience in the partnership side of it. Is, has that come to fruition as an important part of the business?

I would imagine when you're integrating with lots of things, it's this, just to look up the API you do it right, you want to build relationships, partnerships with Other company. So, so we have a very strong ecosystem of partners look really looking to have the conversation. I experienced essentially that assistant experience on top of their platform, and right now we're letting into QA few of the strategic ones that we feel that we can both have a strong alignment with, but also maintain, um, as we grow our team, we're going to have a much larger ecosystem.

We're already at probably 10, 15, uh, very strong players in the industry who want to have a partnership in place. So I feel it's only gonna grow from that point. Good.

Good. I mean, you already know this. I, I, I learned that lesson on one of the products I worked on where you didn't design it with you.

Were gonna have a partner ecosystem in mind, get to a point like, okay, we can't do what we want to do, we gotta go back and add that in. Yeah. So having that perspective front, if that's a place you're gonna want to go at some point in the life cycle.

Yeah. Our ecosystem that, that's exactly it. Our ecosystem is built to be two-way conversation with all tools.

So if our customers are asking for it, there's no reason why the vendor on the other side wouldn't have a stronger alignment. So we can have a, a better user experience for both sides. Mm-Hmm.

I'm curious that, so do you sell to the engineering development kind of community? Do you sell more to the platform engineering, operational DevOps tool kind of folks? Where, where do you tend to focus it?

It's, I would say usually that decision is made jointly. The platform teams typically have to have a handshake with the engineering managers. So whether it's engineering manager taking us to the platform team, platform team taking us to the engineering manager, typically we see both sides involved at least validating this mm-Hmm.

Uh, but the easiest way to get started is really with, um, platform teams because they're the power users typically of the system. They're the ones creating and scaffolding and configurations of the system. So it's easiest for them to get started, test it internally on themselves and very easily.

We usually see that they roll it out to the rest of the engineering team, And they already have a good idea of what kind of requests come their way and Precisely Would make that easier for People. Otherwise, the other way around would be, let me go and bring in the platform team. So it's, it's a more direct approach in that makes Sense.

And, and it's nice. I could totally see, you know, it's starting well, hey, something might make it easier to maybe take a load off the platform engineering team and or of them saying, Hey, we're here to augments easier way to work with us. Right?

We're here to augment them. Not everybody has a luxury of hiring three to five more headcounts for next year, but if you could do so with one or two headcounts and then now you can scale, Great. If you kinda look forward a little bit, maybe this next six months or maybe 12 months, six months or so, kinda what kind of, what things are happening in the industry.

Yeah. Maybe where you're heading. What are you thinking about next?

The most recent announcements by, uh, open app have been very interesting because from our perspective, it's giving us even more, uh, tools in order to get to our end goal, which is, in my opinion, a self-driving operator, self-driving operator that will be able to take end-to-end tasks away from human beings and really free 'em up to do the highest and best use of their time. It's still gonna require human in loop interaction. We don't expect it to be overnight.

Uh, but I think a lot of the things that that was announced recently is already in line with what we've been doing, but it's only making it easier and more, again, unlock more and more use cases. Seems like they, OpenAI have done a pretty decent job of like, don't, don't throw it all out there. Right.

Incrementally add capability Yes. When you're ready to handle it. Right?

Yes. Uh, because we've been in this space a couple years, we've been thinking about, we've already been architecting, uh, our platform to, to be conversational native. Uh, that's allowing us to move very fast in this space.

Great. Well tell folks where, where can they go check it out? Can they get a demo absolute online, or what, what can we do to help?

That's a beautiful part. We, we open our platform up for self-service. So you could actually sign up.

It could be I ai www that could be ii sign up, uh, for a free trial. We also have a sandbox where you could, without needing to put in organizations credentials, anything. It's a gated experience, obviously it's a bit more limited.

Uh, but you could go and, uh, ex have that nice exchange with Kuby, our DevOps assistant and really enjoy the, the experience before committing to going into a trial. Appreciate the sandbox. You don't wanna quite turn it loose on your infrastructure quite yet.

Right. Let's kick the tires and then get There. Mostly read only.

We have a few, uh, ride actions in there, but typically we, we wanna make sure it's, uh, rated PG for what people can do today, So Exactly. Great. And it's spelled K-U-B-I-Y-A-Y-A-I.

Great. Thank you very Much. It was on the, I'm sure you on your title, But yes, I mean, very Pleasure talking with you and congrats on the new progress, new release and capabilities and where you're headed next.

Great market to be in. I appreciate it and I appreciate your time here today. Of Course.

Great to have you on. For our folks, for the people in the audience. So I'm meets representative of some of the great things that are happening not just at KubeCon, but in our industry and, and progress is happening overall.

So stay tuned. We've got more great interviews coming back up. Uh, don't go away.

We'll be right here in a few minutes, so we'll be back.