Elastic AI Assistant and Universal Profiling – Gagan Singh, Elastic

This is Textron tv. Hey everyone. Welcome back here to Textron tv.

I have a person's first time on Textron TV today to introduce you to, I'd like to introduce you to Goggin Singh. Goggin is the VP product marketing over at Elastic Goin welcome to Tech Drunk tv. Thanks, Alan.

Good to be here. Nice to have you on. So I mentioned your VP product marketing at Elastic, but goin share with our audience, if you don't mind, a little bit of your background.

Yeah, Sure. Alan. Uh, so I started off as an engineer actually, and so I've been as an, uh, I've been an engineer in my, uh, for a number of years in my life, and then made the transition over to the product management, product marketing side of the house.

And I have been involved with the, uh, observability world for a number of years, working for different companies across the, across, uh, you know, the industry. So I've been really familiar with how this space has evolved and from the monitoring days into the observability and, and further on. So, um, it's, so I've seen it firsthand in terms of what the evolution has been, and it's been exciting to see it.

Very cool. Very cool. Um, and Gagan, I think most of our audience is familiar with Elastic, though a lot of it may be from Elasticsearch or stuff like that, but, um, just in case to clear up any misnomers or people who aren't, why don't you give us a little elastic kinda 1 0 1?

Sure. Happy to do that. Yeah.

So as, as you said, a lot of, uh, uh, folks know us for Elasticsearch. We've been around for 12 years. Uh, and actually what Elastic is, is, uh, you know, we've, we've over 50% of the Fortune 500 companies that are out there, and our intent really is to democratize search.

That's how Elastic started off with, uh, and now it's the ability to use generative AI and bringing the power of AI to make sure that, you know, customers are able to find the answers from what they need. Uh, and in general, the way I would categorize Elastic or Simpl, simply say that is Elastic helps you find answers to the data that matters, uh, in real time and at scale. And built on the Elastic Search platform are three solutions, which is basically observability, security and search, which are, you know, really adopted by customers.

For example, for observability, we have full stack observability capabilities that are available within Elastic. Excellent, excellent. You know, look, I've, I've, I, I actually, I think we might use Elasticsearch on all of our sites as part of the, uh, WordPress stuff, but you know, to me, elastic was always about scalability, right?

The ability, if you have large amounts of data that you need access to and, and, and make actionable intelligence from, whether it's search or as you say, observability security. The elastic was, you know, when we talk about big data solutions, and I may be dating myself by saying that, right? But there was a time where it was quite a feat to get your arms around truly big data.

We do it better today than we did 12, 15 years ago, right? But it, you know, it's, it, it it's quite a, it's not easy, right? It's not the kind of thing, you know, there are a handful of companies that do it well in elastic's, one of them.

Um, now of course, everybody today is talking about AI generated of ai and you know, there, there's, I've seen, I've seen arguments back and forth that the bigger the l l m the better your AI is. Of course, the bigger your l l m is, the more poisoning or hallucinations you could get, right. You know, so there's gotta be a, a sweet spot, I guess, in the middle.

But I would imagine Elastic would be, you know, right dab in the, in the thick of it when we come to generative AI and LLMs with its ability to, to, you know, wrap around that kind of data. Yeah. Yeah.

And thanks Ellen from, uh, for, uh, talking about that because one of the ways I would say is that, as you've mentioned, right, there's, uh, there's been a lot of challenges in trying to manage data, but that data explosion has not reduced. It's actually in fact, grown exponentially. If you look at it from a customer standpoint, there's structured unstructured data that's growing, uh, really exponentially and very rapidly, and that makes it a big challenge for customers.

And I think that's really this, as you mentioned, the sweet spot for Elastic has been, we've been working on democratizing search and making sure that it's available to everybody. We've been working on ai ML capabilities, vector databases, and so on. And so the way we are looking at it is to say, you know, generative AI has obviously taken it to the next level where, you know, it's really about the interactive natural language sort of, uh, mechanisms that people can interact with the, with the AI capabilities.

And that's what makes it really, really powerful. And, and yes, we have been in the thick of it. And so, you know, one of the things, for example, we've been working on and we've, um, really recently announced is for example, the AI assistant.

We have AI assistant for Elastic observability, which we'll, we'll talk about. And, and the whole concept was really driven from the fact that if you look at something like observability the solution, you've got various types of data that are coming in, whether it's metrics, logs, traces, you may have heard of that, as well as profiling data, which I'll talk about as well. And then, you know, also important is the business data that goes along with it.

Now, uh, a lot of times, you know, in, let's say, you know, in the data center world, you were looking at obviously trying to solve this problem in silos where everybody had some tools and they were all just running around trying to find, solve problems manually. You open up a bridge, right? And then you bring experts together.

And so obviously the industry gravitated more towards observability where, because the boundaries as the cloud came in has been, you know, sort of dissolving away. Uh, you know, there's, uh, and, and there's less and less of these different silos, but what still has continued to happen is there's a lot of data which people need to make sense of. And so with the AI assistant, which we are just announcing, uh, what is, what it really does is it helps you bring these silos together.

So, uh, you are able to interact with that data. You are able, you don't need expertise in any of these areas. So what it does is it reduces the learning curve for any types of operations teams, ss r e teams and developers and so on, as well as just making sure everything is context aware.

Um, and I'll, and I'm happy to go a little bit deeper into some essen an essential thing you touched upon, which is really around hallucinations and the need for, you know, bringing in your own private data, uh, with the LLMs to make sure things are context aware. So happy to jump into that if that's, if that, that's helpful here. Yeah, No, you know what, that's a, a great topic, Gargan, uh, because we had a hackathon here, oh, almost a month ago now around operationalizing ai, right?

And, and this whole concept of people, you know, taking additional data, let's say into a Vector database today and, and injecting it into an L L M, right? Um, I, I think that's becoming sort of the preferred method, if you will, of people A, trying negate hallucinations in poisoning, but b, also making, uh, our AI functionality more context aware of our individual corporate organizational needs. Right?

Right. Yeah. And, and that's, I think that's a very critical point because, you know, you can obviously from, let's say an AI assistant or any sort of a copilot perspective, you can connect into any of the Open N LMS that are out there, the open ais of the world and the Azure Open ais and others of the world.

But, uh, what that does give you is basically just generic information and that leads to hallucination. So for example, if I am in the operations team in let's say enterprise A, what does that really mean for me? Yes, sure.

Maybe there's some generic responses that may or may not apply to me. Uh, and, and, and to, to your point, what that needs is, you need some sort of a solution that is able to give context to that private data. And, you know, as you mentioned, elastic has been working on the Vector databases, and we had actually also announced something called Elastic Search Relevance Engine, which is sra.

And what it does really do is it actually bridges this private data as well as the public data. So it's actually continuously learning, using your private data to be able to provide any responses to you which are very context aware, so that it understands what your, you know, enterprise is interested in. Uh, or for example, if you have Runbooks or any sort of automation, uh, capabilities and techniques, it is actually able to suggest those to you, which are very, very specific to your organization and how it's set up.

And so that's, I think that's, as you know, as users, as customers are looking for it, I think that is an essential aspect that they have to consider as, as, as they think through this, you know, the AI technology and we see customers who are building their own LLMs, we are able to, you know, add, uh, information to those LLMs through all the learnings we have using private data. So we're really able to do some very cool, uh, context of their insights that, that are highly, highly actionable by the user Fair. Absolutely.

Um, so the, the AI assistant, which, which Elastic has, uh, announced though is specific to the observability? Uh, well, I mean, so we have last assistant for observability. We have, we have also launched it and has been in the market for a while for security as well.

So ultimately, oh, okay. I didn't know where I wasn't aware. So Ultimately, the intent for us is, uh, is for, for us to make sure that any of the operators and operations teams or the developers who are using the Elastic platform are able to get the capabilities of AI assistant, and it's embedded throughout the platform, so as well as the solutions, right?

So that's how Elastic is, is we make sure that we are democratizing everything. There's nothing that, you know, stops you or prevents you from using these technologies, and they're all embedded throughout, for example, observability as well as the security solutions that we have. Excellent.

Excellent. If you don't mind, uh, going, I, I'd like to turn, there was another, uh, piece of it in the announcement and on that topic for today, and that is universal profiling. Yeah.

Yeah. And so I think the way to think about universal profiling is that, you know, profiling, uh, from a traditional a P m application profiling standpoint was a fairly heavy mechanism to be able to profile or do anything and get insights. And so as a result, you know, uh, the development and operations teams had to instrument that code properly and they had to restart some services and things of that sort.

And doing all of that historically in production is really, really challenging. And as you can imagine, it's impactful as well because it may impact services, it may impact the, uh, performance of the services and the applications involved. And so what we have announced is the general availability of, uh, universal profiling, uh, which is basically, um, you know, is, is a continuous profiling technique that leverages E B P F capabilities to give you really deep insights into your application as well as infrastructure.

And the cool thing about all of this is that we are able to give you this continuous profile information with very low overhead. So the overhead in terms of memory and C P U is really, really low. And we've actually had customers who've achieved at least 30 to 35% savings.

And, you know, that helps them go green too, because as you can imagine, there's a lot of focus in making sure that your compute and, you know, the, the footprint that you have is as optimized as possible. And, um, and so that's something that we're really excited about. Uh, the other cool thing about it is that we feel that this actually compliments the application performance management where, you know, you obviously need a lot of, um, uh, instrumentation to be able to use distributed tracing to be able to actually trace, you know, what the problems are in the code, but that requires multiple teams that requires the code to be instrumented.

This is very simple where any operations or SS r e team can drop in the universal profiling and be able to profile the application or the infrastructure fairly deeply without, you know, needing additional teams help. So it actually helps 'em at least pinpoint the areas and we, so we feel it's really, really complimentary, um, and, and without any impact to the performance. Excellent, excellent.

Gargan, I, I don't mean to put you on the spot here, but you know, elastic is a company that's known to play very well and is a big player in the open source community, um, making their products available to a lot of people, you know, free as in beer as they say, versus Freedom. Um, I, I, we, you know, both AI assist and for both for observability and security as well as universal profiling. How are these offered?

Are there free versions? Is it strictly a pay version? If so, like what, right.

What does pricing look like or give us some idea? Yeah, I mean, so, so both, so both these capabilities, you know, as is with Elastic, we are, um, these are all available to all the customers, so they can actually try out today. Uh, AI and AI system for Observability, for example, is in tech preview.

So they can all, they can try it out. And these are all available as part of the free trial, which the customers can use, uh, and and, and actually get, uh, get going. And so there's no additional pricing.

So for example, this is where we are really, really different from every other vendor in the space where you don't have to continue to pay for additional functionality. You just, uh, you know, you are getting these, the moment you sign up for Elastic, you have access to all the elastic capabilities they get of these solutions as well as the platform. And I think if I may, Alan, just one other thing I'll just add here is that we to the point of Open and, uh, free, you know, you talked to Connection a while back where we had contributed Elastic common schema to Open Telemetry, and we continue to, uh, you know, focus on Open Telemetry as the data of CHO at the architecture of choice, uh, and, you know, make sure that customers have the flexibility and are actually have the freedom to, to adopt the vendor or to the capabilities and technologies that are, uh, that are of interest to them.

Absolutely. And, and, and thank you for that, you know, not you personally, but the whole elastic team. Gagan, we're about outta time.

I want to thank you for coming on here your very first time here on Text Drunk tv. I hope it won't be your last. Um, keep up the great work.

You know, it's exciting to see how AI is, uh, putting its mark on so much of, of what we're doing out in, in, in the operational IT space. And, um, and this is another great example of it, you know, what we forgot to mention of all things. Where can people go get more information about both of this?

Yeah. co site and you'll see the information about it. We're gonna be announcing it to the market.

So really excited to see, um, to be talking about it. Great. And, and thanks for the opportunity, Alan.

I really excited. Really. My pleasure.

It's great to have you on. Thank you. Goggin Singh, VP Product Marketing Elastic here on Tech Drunk tv.

We're gonna take a break. We'll be back in a moment.