Causely Brings Reliability Engineering to the Heart of Cloud-Native Development with Yotam Yemini

Hey everyone. Welcome back. We're live here at CubeCon.

You know, I guess the, the, uh, show floor is open 'cause I see a lot of people walking around, but it is a huge show floor and we're down. Well, I, you are looking this way, but behind us that way is the meals and seating area to our right is all of the, uh, different projects from the CNCF. I think there's over 200 projects now, right?

Yeah. Crazy. Um, and it's getting a little crowded here, which is good because it brings the heat of the place up.

It's so cold down here. Um, let me introduce you to my next guest. His name is Yotam Yamini.

Nailed it. Did I get that right? A Plus.

Beautiful. Yo Tom is the CEO of Causley company. You probably heard about or you've heard of, but I don't know if you know everything they do.

Yo, Tom, welcome. Welcome to Tech Drunk tv. Let's start with you.

Yeah. You're the CEO of Causley. How did, how did you get here?

So, Causley's, my fifth startup, the first one that I'm fortunate enough to be the CEO of, uh, but actually this whole team that I joined here at Causley was a team that was part of my first startup, a company that was called VM Turbo. Okay. Remember VM Turbo?

Mm-hmm. Eventually, we rebranded the company into Tur Omic Uhhuh, acquired by IBM for $2 billion in 2021. Beautiful.

And in 2022, Causley was founded. I actually joined the team in 2024. At that time, I had actually gone to another startup that was acquired by Cisco.

And so I was working inside of Cisco, running the incubation team for the security business unit. Okay. And Schmuel, who's the founder of Causley, who I believe you've had on your show, uh, pulled me out of, uh, Cisco and said, Hey, why don't you come join us here in Causley?

And, uh, for me it's, uh, it's like a full 360 like, you know, career moment to come back with the team where I started it all To your team. That's a great thing. I, you know, look, this is tech strung is probably the fifth or sixth.

Yeah. Yeah. At least startup.

I've done two though. This one. Not venture back.

This is the first one I did that, you know, new venture and I, I, there's something to be said for working with people you're comfortable with. A Hundred percent. You know, you need fresh blood, but when you work with people that you've worked with, in my case 20 years Yeah.

Or more. Yes. Especially remember my exec team.

Yes. It's, uh, it's comfortable. Yeah.

Well, especially if you're trust, if you're a remote first company like we are. Oh, yeah. I think that added measure of trust and knowing how to communicate with each other.

Yeah. You know, when you, you gotta kind of get right to the point, uh, really helped. Huge.

Huge. And, um, so did you come in as CEO in 2024? Is that it?

I did. Yes, I did. Very cool.

Um, so I mentioned a lot of people out here may have heard of Causley. I don't know how many of them really know Causley though. So if you wouldn't mind, yo Tom, let's start there.

What, what's Causley? Well, you in your own Words. Yeah.

So we're on a mission to help people automate service reliability. Uh, we can unpack that further. But, uh, if you had schul in this seat, he would tell you that this is really a continuation of his life's work.

So the first startup that Schul co-founded was a company called Smarts in the nineties. Yep. And, uh, smarts was a leading provider for root cause analysis for networks.

Uh, Turbonomic just jumping ahead to the second company, that's where I joined the team. And Uhhuh started working with Schul very early on in the day, uh, was automating resource management for cloud and virtualized infrastructure. If you think about this, this is like the continuation of that journey, which makes a lot of sense because cloud native applications are an interconnected web of microservices that are very functionally similar to complex networks.

Sure. Uh, and what everyone is trying to do is trying to apply AI in a way to automate the reliability of those services. And, uh, we believe we have a unique insight and, uh, a lot of lived experience to solve hard problems in the space.

So let me ask you an honest question. So I I, I remember talking with Sri, he lives in New York. No.

Yes. Um, brilliant guy, right? Yes.

Yeah. He's been working on this issue long before AI was a thing. Yes.

As he likes to say, the first hype cycle of AI in Katie, right? Yeah. The first time.

Yeah. It's pretty funny. Um, but you know, today it's disrupting everything, right?

It's, it's, it's sucks the oxygen from everywhere. Um, how has, I mean, and in some ways I guess it's been good for Causley 'cause you are already on that road for this. Yes.

But in some ways it had to have disrupted Causley go to market, right? Because now you've got to accelerate AI adoption, accelerate integration, accelerate, you know, riding this wave. Well, I think in, in some ways it's been a good tailwind.

I mean, I, I think you've heard this term AI Probably the best tailwind you've ever had. Yeah. You've heard this term ai, SRE flying around.

Yep. Uh, and so there's a lot we can unpack there. Uh, but I think if you go back to like our time at Turbonomic, we used to tell people we were gonna help them automate where to place workloads across infrastructure.

And the idea of automation was scary to people. I think now, I think now we're sort of seeing this turn where everyone's saying, okay, I have to find a way to drive productivity gains. I have to use AI to automate things.

And, uh, and the good news is we're sort of at this like nice intersection where people are trying to automate operations and they're learning what they can do with these agentic systems. And also they're learning what the limitations are of those systems. And that's where we really have a unique value proposition offered to the market in terms of what our, uh, product does.

Excellent. Love it. Um, A-I-S-R-E.

Let's unpack that one. Yes, that's a good one. So like it, thank you so much.

You know, like, like every other job function, I think in tech, AI is having its way, if you will. I don't know if it's patently obvious to our audience how this is applying in in the world of SRE. Yes.

Let, let's talk about that a little bit. Sure. So we'll talk about what's out there and we can also then sort of unpack some of the limitations.

So some of the good in the space is it makes a lot of sense that if I need to write a postmortem and learn from why an incident happened, using a language model to write up a summary makes a lot of sense. Yeah. Uh, it also would make sense that if I'm trying to parse through a lot of unstructured log data to look for something meaningful, again, using a language model, which in a lot of ways is like a pattern matcher or a pattern finder could make a lot of sense.

And so I think what you're seeing on the market today that's gaining some traction are these effectively like new age chatbots that SREs are using to help them troubleshoot. Uh, Gartner recently came out with a cool vendor report around the A-I-S-R-E space. Really, they, they missed us on that one, but they're gonna get us on another one.

But one of the things that they, uh, mentioned in that, uh, in that article was that every A-I-S-R-E company they covered is missing the point and missing the fundamental opportunity, which is that the bigger opportunity is to actually help the service owning product development teams to automate reliability as part of how they build their products. And the idea to only apply AI to the SRE function as an SRE function is actually missing this opportunity of bringing reliability engineering as a sort of first class citizen to a product development team. So I have my own opinions with Gartner, but I'm not gonna voice them here.

Uh, over the years at other startups I've done, of course I've paid the money that I think we all pay in ransom to guard If there is a tax. Yes. Um, We have not paid the tax, we were not in the report.

Right. The original ransomware. But, um, let's talk about this though.

If we, if we use AI to build that automation and reliability at the product level, doesn't that make sort of SREs a little obsolete? So it's an interesting point. I would tell you the best SREs I've ever worked with always describe their job as automating themselves out of a job.

Sort of like that commercial, I want to be obsolete by the time I'm 35 or Whatever. Yeah. Yeah.

And you know, you've probably heard the saying like, AI's not taking your job. The person, someone who knows Ai Yeah. Someone Who knows how to use AI to do your job is taking your job.

Yep. Uh, but I think there's, look, I I, AI at this point for a lot of people just means using language models, which is different than what Causley is doing. And I think that starts to lead to the limitations of some of these AI SRE that are out there.

Uh, and people are starting to learn that, you know, it's like cha pt, sometimes it's right, sometimes it's not. And you know, if you're trying to automate reliability, you can't use something That's Right. Coin flip, whether it's gonna be right or not.

Yeah. I used, so I was in the hosting business for a long time and back then, and even still today, you know, five nines give you an SLA five nights. Yep.

But what a lot of people didn't realize is even with five nines, I think it's four minutes or so, I think I how many minutes a month? Yeah. You're down and you're still in your five nine SLA And for people who are mission critical, those couple of minutes a month means a lot of money.

Yeah. Well, look, I'll go even the next step, which is that, and again, maybe this is, uh, a colloquial thing that people already have grown tired of talking about, but for some people, slow is the new down, right? Yeah.

So there's also the point that if you can apply the ethos of reliability engineering, and you take SRE in terms of the idea of error budgets and SLOs for every service and a blameless postmortem and all these other sort of ethos elements of the SRE culture, and you apply that to how you build software, you end up not just troubleshooting things faster, but preventing issues in the first place. And that's really the big opportunity that we see, particularly for how our model and our system works. Love it.

Now, so do you think you are, who, who's your target customer here then? The product manager or the SRE? So, uh, I'll give you a different persona, which is the software engineering manager.

Okay. So the people that are building the actual customer facing services and applications, uh, we have real money gaming clients, uh, that they use us for their sportsbook app and their casino app. Sure.

Uh, we have SaaS companies where, you know, it's the development teams that are building the customer facing product. Uh, and look, at the end of the day, we also do have SREs and platform engineers that use the product. No, I was just gonna say Plat, the software engineering teams, probably a lot of platform engineering in here now though too.

Yeah. Because A lot of the shape of these problems with cloud native applications is, um, because everything's an interconnected web and you're trying to figure out is the problem in the app or is it in the infrastructure? Or is it one of my Oh, yeah, yeah.

One of my, uh, friends likes to say, is it an issue or an ish? Me? Yeah.

Right. Yeah. It's, and so, and so, you know, it's, it's a problem that really spans personas and user types.

Uh, but we really do find that when we are the most successful in a customer, it's because an engineering manager or director of software engineering really adopts the idea that, Hey, I wanna automate reliability work. Help me do that. So it's, it's, that's not really, that's not really top down though.

It's not. 'cause it's not the same middle. It's like middle out.

It's Middle out. Yeah. Up and down.

Yes, exactly. That makes sense. And then, but man, we jumped into things and I didn't get a chance to talk more about Causley the website.

Yes. ai. com.

Sharks love sugar. Love it. Yes.

So now you've said it. How, where's that come from? Well, so you know, the, the sort of next best alternative to what causally does, which is we're building this causal model of, uh, the and effect between things in your environment.

The next best alternative is the correlation engine. And so we like to sort of have this play on correlation is not causation. You know, just because there are more shark attacks in the summer doesn't mean that sharks love sugar sugar.

Right. Because, you know, ice cream consumption also goes up the summer. Got It.

The idea. So that's sweeter. Yes.

ai. It is, um, I know you guys recently announced an MCP server now without making it a Me too. Tell me what's the deal there?

Yeah, So look for development teams, which is who we're really trying to appeal to, you gotta be where they work. And where they work is in their IDE. Yeah.

And so everything that Causley does, we expose not just through our ui, but also via an API also via webhook. And now, as of recently, an MCP server as well. So if you're a developer and you're in your IDE and you, for example, want to, uh, improve the Java heap size, uh, in a, in A-J-V-M-H, how do you know when it, when you're actually gonna get your bang for your buck to do those types of things?

Uh, or let's say you want to change, um, the, you've got a slow database query that you need to optimize. You could run a profiler to always do those things, but how do you know you're really getting the juice for the squeeze? Right.

And so our causal model is helping you understand the actual cause and effect in terms of sort of at a meta level, which optimization actually gets me something from a performance perspective. You know, sort of the bank for the buck of the work I'm gonna go do. So we're exposing everything from our causal model now via an MCP server.

I love it. I love it. Um, you know, isn't it amazing though?

I mean, it's, it's, I think to be fair, I think MCP came out last December, not January or February, but within a year. I mean, we talk about it like it's, you know, table stakes I emptied up, right? Yeah.

Let's play poker. Um, at some level, I, it still kind of freaks me out how quickly this whole thing is really just coming together. Um, well, I think, Can I add a little thought on that?

That is that I think we're seeing like markets go in waves from bundling to unbundling. And I think that particularly in the part of the market that we're in, there's this sort of unbundling to build your sort of best in class Right. Observability stack.

That that's always how that pendulum starts, right? Yeah. Especially when it's newer stuff because it's, we're still figuring it out.

Okay. And then once kind of best practices really emerge, then you go back to the buny at some point. Makes the world go round my friend, right?

Yes. Um, what about Kup Con? Yes, but so far so good.

It is buzzing, like you said, it's starting to buzz and, uh, I'm excited about what's here. It's a very, um, it's a very passionate crowd. I, I will tell you, you know, I've been covering Kup Con now for, I don't think the first year, but eight or nine years.

And, um, it used to be all developers and all people like in t-shirts and backpacks over the LA and I do both here in the European one. Over the years I've seen more people wearing sports jackets, uh, more ops people, of course SRE people, platform engineers. So it's diversified into, uh, quite a community.

But it's a pa a community that's passionate, it's passionate about open source, it's passionate, obviously, about, you know, doing more faster. Yeah. In the cloud though.

The other thing about cloud native is it's not necessarily in the cloud. Yeah. Yes.

Every everywhere. Yes. It could be on bare metal.

It's on the edge. It's everywhere. Anyway.

Hey man. Pleasure having you on. Hey, great.

Thanks for having me say hello to Spiel for me. Yes, thanks. Um, let's see if I remember this right.

Yo, Tom, y yai yamini yamini. You gotta agree it the first time I did. But just flip it.

And if we weren't live, they could it and replace Oh no. They, they expect this from, you know what I mean? I can never do it.

I'm like, ai, I can never do the same thing twice. Exactly. com.

Yes. Very good. Thanks again.

I love it. All right. We're live at Cuon.

We'll be back here in a moment.