Observability Rebuilt for the AI Era

Hey, everyone. Welcome back here to Techstrong TV. I want to introduce you to my next guest.

It's his first time here on Techstrong TV, hopefully not his last. His name is Gabriel James Safar. People call him GJ, and that's what I'm going to use.

GJ, welcome to Techstrong TV. It's great to have you on here. Thanks for having me.

Well, it's our pleasure. I'm going to start off with something real basic. Give us the deal with Gabriel James.

Ah. The story comes from the fact that my mother wanted me to have my grandfather's name somehow, and she wanted me to have the same initials, anagram, as my grandmother. And so my grandfather was named James, and my grandmother's first name started with a G.

A G. And so GJ. Voila.

I got it. That's a great name. Good for you.

My youngest son, so we have Hebrew names and English names, right? And so my youngest son, there was a lot of people to name him for, and he has a similar kind of thing. He has about three, four different names in his thing.

So I get it, believe me. But anyway, let's talk a little bit. So you were born, they named you Gabriel James, your friends call you GJ, but what happened then?

How did you wind up as the co-founder and CEO of this company called Suga? Well, like many people, I studied engineering, became a data scientist. Started working as a consultant in data science for a few years here in Paris.

I'm born and raised here. Uh-huh. And then I was bored, and so I fetched a friend from university, and together we created a first company.

Total failure. Was a terrible idea, but at least we tried, right? Uh-huh.

That's how you learn. Yeah, exactly. And then we created another company out of a problem that we had faced during the development of the first one, and that company became a test automation product that got acquired by Datadog after a year and a half.

So that's how we joined the observability space and Datadog, after they proposed us to join them to create a suite of products. And we had an amazing time there. Got very lucky, could build a lot of stuff and, after a while, wanted to try to find the meaning of life.

So independently, we both took a year off, a little bit more for him, a little bit less for me. Failed again at finding the meaning of life. Did not find the meaning of life.

You're not the first one or the last. It's okay. It's okay.

At least we tried, right? Just like- Uh-huh ... the company.

At least we tried. I get it. I've learned.

Look, I have a boat, and when I first got my boat, if I was going to point A, I would try to get to point A as fast as I could. Somewhere along the line, I learned it's not about getting to point A as fast as you can. It's the journey to point A.

And that's not the whole meaning of life, but there's a lesson there. And so even though you may not have found the meaning of life, the fact that you took that journey to find it, it was in and of itself something. Yeah, for sure.

And in the process, I had a second kid, which was also a good thing to spend some time on, too, right? Beautiful. Yeah.

And then, after that time, we were ideating about, okay, we have energy. We want to build stuff. What are the problems that we faced?

And it was the very beginning of the new AI era, right? Mm-hmm. And so we were thinking about, okay, what will it do?

What will be the impact of that era? It was not entirely clear that it would have an impact on observability, the space that we knew the most, having spent a lot of time there. But what was clear is that as the people in charge of talking to the largest European customers for Datadog for a while, we saw clearly that there was a gap in the market for these companies.

These companies cared for governance, they cared for sovereignty, they cared for scale, and it was not something they found in the products in the market. And so we felt, okay, we know the space pretty well. We know that there is a need for a variety of companies that suffer and would like products to fit their needs.

And that's how we decided to create Suga. And then, now two years later, AI came on, and the problem became even more acute for most of these companies. And so here we are.

Sure. Sometimes timing is everything, right? You play into a wave, and as that wave crests, you're at the top of the wave, if your timing is right.

Sometimes your timing's wrong, and that wave is crashing, right? And it happens, too. What a great story, though, right?

And it's you and the same co-founder who's now on third go-round here. Yeah. Now he's the godfather of my son.

A very dear friend. Good for you guys Yeah. We've known each other for 14 years, yeah, now.

So it's been a while. Third company together, so yeah. Good for you.

Good for you. Let's dig in a little to Tsuga, though, right? You looked at this...

Observability is huge market, right? Because it really kind of evolved out of what we used to call application management, process management, APM, or whatever you want to call it. And then built on top of that, and then of course, open telemetry and the open source, the cloud native kind of products that helped feed into it, have really made observability huge today.

But really at the heart of it, though, is our ability to wrap our heads around, or at least wrap our programs around huge data sets. Because to really value, to get the value out of observability, you need to be able to observe many different things. Now, when we say Tsuga is AI native resilient observability, talk to us about what we mean there.

Okay. So maybe the first piece is what is a tsuga? Sure.

A tsuga is a family of pine trees that grow out of Japan and in the Seattle region, I think. Really? Okay.

Yeah. And one, my wife is Japanese, so, I liked that small- Sure ... small gesture towards that.

But secondly, it's a tree that has good properties for building things. " It was really, how can we tailor each platform, each observability platform for the needs of the customers to fit with what they need? And so, to do that, first we have our deployment model.

The deployment model is bring your own cloud. So we deploy our product onto their infrastructure, meaning that the data doesn't leave their premises, meaning that they can choose, in terms of data residency, where they want to put their data. Meaning that in terms of compliance, we don't touch that, so they don't leave the control to another vendor.

Sure. They are not abandoning the value of this data to a third party that could exploit that, in most cases, to develop models and learn stuff and resell it to other companies, right? Instead, we are compatible with any open source data collector, most notably OpenTelemetry, of course.

We help customers deploy these collectors so that they have the best setup. So to optimize the volume, the completeness of data collection, et cetera. With our forward deployed engineers.

And then the data doesn't leave their premises, and then it's stored in their infra, with a very good compression rate, so that in the end, they can stop caring about the retention of their data. So they can really do whatever they want. The collection is open source, the storage is open source in their system.

It's compatible with all the BI suites if they want to consume the telemetry. For example, to merge it with the business data. So all the Arrow, Data Fusion, Ballista, all these suite of tools, the storage is compatible.

So there is no lock-in, right? The collection is theirs, the storage is theirs, the infra is theirs. We are just here to provide the right piece for them to make it work the way they want.

Good. Excellent. Now, you know what, GJ?

We didn't even mention you guys recently also did a Series A round? We did. Look, our audience is generally not the finance team, but might as well mention it and give us some details with it.

Yeah, of course. So, we had started the project raising funds from General Catalyst, and a few other investors out of Europe. And after a year and a half, we went out of stealth.

We announced the product existed, we had a few design partners. And then the sales motion was quite successful in the following month. And so a few months after going out of stealth, we had enough success to motivate a few investors to join us.

Not only to bulk up the GTM, which is of course- As you know it ... yeah, of course, it's very important in the end, notably because we want to serve the largest companies, so we need to ensure that we provide them with the right people, to support them, help them on the project management, on the deployment, et cetera. But also to accelerate the development of the platform, and make it even better.

So, we found the right partners, our existing investors, so General Catalyst, Singular, a French VC firm, DST, joined in the round. And we are now recruiting as fast as possible to- Yeah ... to develop the product and make it even better.

Notably because the way we see observability is that, as you mentioned, a few years ago, there were APM, infrastructure monitoring and metrics, log management. There were three different products. In the past eight years, it's been obvious that it's just one category.

Observability. Yeah. Observability.

And actually, we think it goes even further, and user sessions, net flow logs, CI efficiency profiles, all these elements that help you understand how your system behaves are actually just one problem named observability. And so we believe that it's very important to support all these data sources, but also to provide catered views to enable users and AIs to understand these elements, thanks to this telemetry as well as possible. And so, that's what we are building now with that.

So improving the product, making it broader, and that's super exciting. Absolutely. And I just want to touch on one other thing, and because I'm in the middle of doing a large report on sort of, I call it the grand unification theory, how development, DevOps, platform engineering, cloud native, all of these disciplines, if you will, all of these foundational models, whatever you want to call them, are being brought together now with AI, right?

Because AI is drawing them all into, just as you said, how observability is being drawn into just one category. I think we're seeing software engineering, right, taking all these disciplines in and making them more tighter. Go ahead.

Yeah, sorry. I would go in your direction. I think what's happening at least is we used to have a lot of verticals- Yeah ...

and I don't know if it's going to become one vertical, but for sure the verticals are, let's say, aggregating. Yeah. So observability is for sure one vertical.

BI is of course one vertical. Software development life cycle, right? Everything around CI and artifact management, and that's becoming a vertical.

So that's clear that at least everybody sees the value of these aggregates, and I think it's because software development goes so fast now that having value, while switching tools is becoming more and more complex. And also, one thing that we clearly see in observability is if you have not designed your tool for what's happening now, it's very easy to become obsolete. A very good example, I think, is we provide our customers with an AI SRE, with the model of bring your own agent.

So instead of meddling with their AI policy, which can be tricky, we tell them, "Okay, whatever agent you validated, you can just use it to- Mm-hmm ... " AI series can perform hundreds of queries at once. And customers that stayed on older storage systems for their telemetry are just crashing the observability platform, just thanks to one RCA agent performing an analysis.

So to me, what the AI era is bringing also is the need for extremely resilient technologies that can support agents performing tasks 100 times faster or 100 times more aggressively than human beings without breaking. And actually, they should be built to help these agents, like aggregate the information so that the token count is lower to perform their actions, or provide aggregates so that they can explore a smaller space instead of a bigger one. And so all these things that we are embedding in the platform because we are built in that era, is something that is probably absolutely needed if you want to keep up with the pace of software development these days.

Agreed with you. GJ, I promised you this is only a 15-minute interview, but we're way over that. I apologize.

We've got to end it here. But you know what? Come back on.

We'll talk more in the future. I'm sure you'll have more news soon. In the meantime, though, I want to wish you and your co-founder and the whole company success with Tsuga.

We'll be watching it closely. It's going to be an interesting ride because we're in interesting times, as we said. You know what we didn't mention?

What's the website? com. Very simple.

T-S-U-G-A. com. Exactly.

Love it. All right. Thank you and best of luck.

GJ Safar, co-founder, CEO of Tsuga, here on Techstrong TV. We're going to take a break. We'll be back with more.