AI’s Evolution with Abhinav Puri at SUSECON 2025

This is Textron tv. Hi everyone. It's Alan Shimmel here back at Scon 25 in Orlando.

Uh, if you've stayed with us and watched a whole bunch of these videos, this is the last one we were recorded. It's kind of bittersweet. We've had a great time in Orlando.

It's been a great Susie Sussan. Uh, if, if this is the only video you've seen from Sussan, you're missing out, go check out the rest of our videos of an article up on Cloud native now as well to check out. But let me introduce you to our guest.

We kind of maybe save the best for last. I want to introduce you to Abinov Pre. That's correct.

Okay. Abinov. Uh, well, Abinov, first of all, welcome.

Thank You. Thank, thank you for having Me out. My pleasure.

Why don't I did, I did my piece. Oh, yes. Now it's up to you.

Introduce yourself to the audience. Tell them a little bit about you for Sure. So, hi, I'm Avinov Perri as Alan.

She, I'm the general manager of for Portfolio Solutions and AI here at suse. I've been with suse closing in on a decade now, and I've worked across the entire portfolio from systems management to the cloud native portfolio, to doing AI now in the last couple of years. So I've seen everything.

I've seen the evolution, and now I'm seeing the revolution and transformation with ai. Absolutely. Well, you just said it, right?

What a great time to be involved in ai. You bet. It's going on.

Um, you know, we've talked about so much here this week, the DevX, uh, frameworks, the, the partnerships with, you know, companies like Dell and HPE and AWS and, you know, so much going on in, in, uh, in the, uh, VM space and everything. But we haven't really, we've touched on ai, but we haven't kind of, I think, bit the main mm-hmm. Piece of it.

And that's why I'm glad you're here. As you look out at s suse today and where you think it's heading in the short term. Yeah.

Who knows what the long term holds. Yeah. What role is AI playing.

Mm-hmm. And, and what kind of ai, 'cause you know, AI comes in a lot of flavors, That's for sure. That's for sure.

Mm-hmm. So, let's talk about, you know, I'll share where SUSE is going to evolve in its journey with its portfolio in the space of ai and how it will help organizations leverage the powerful platforms, the tools that they have to drive applications in the environments. So at suse, we are positioning ourselves as a platform for enterprise organizations to run their AI applications on top of.

So we provide the foundational infrastructure that organizations can leverage for any gen AI application or other AI applications that they want to run. We provide them the tools, the technologies, the curated and secure frameworks that they will need to build their ai. And then based on the SUSE Rancho Prime platform, they can deploy and run their applications, manage their entire cloud native AI workloads to their entirety.

Now, does that mean it's optimized to run our GPUs? Yes, for sure. We, with suse ai, we can leverage the most powerful GPUs out there.

So let me share a bit about, uh, you know, what SUSE AI is doing and all the interesting things that, uh, we are doing with suse ai, for that matter, the world of ai, as you, you know, you rightly see and you see around you, it's rapidly evolving. New tools, technologies, frameworks, they're launching every other day. What is relevant today, what you and I are talking about today may not be so relevant three to six now.

So three, who knows, right? So, in such a scenario when organizations are building out their AI platforms, how do they ensure that they can keep up with the pace of innovation in a trusted and secure manner that they can innovate and build their AI platforms and to help them do that? What SUSE is doing is that we are curating a library, what we are calling the suse AI library of open source tools and technologies that SUSE is investing to curate and secure for our customers across this varied AI ecosystem so that they can innovate faster.

And not just that we provide them the suse AI library competence, and then we also provide them the blueprints, which are step-by-step implementation guides so that they get faster time to value with those components. So they leverage their investments fully. So that's one thing that we are doing with ai, but we're doing more.

So think of those scenarios. We all work in different types of organizations. In most organizations.

We have these complex multi-step business processes and workflows that involve some sort of data processing, some sort of decision making, some sort of human intervention from people like you and me to solve something. They're ripe with inefficiencies. We all have these inefficient processes in our organizations.

Mm-hmm. They hinder any organization's ability to move forward at pace. This is where a concept, and you mentioned that while you were talking off camera of AI agents Yeah.

Comes into play Now, AI agents help automate a lot of these repetitive tasks. Yes. Like, think of a scenario, let's say in healthcare, right?

Where you are talking to an AI application and it is helping you diagnose a possible illness. You're talking to it, it's reasoning or by itself. Mm-hmm.

It's asking you intelligent follow-up questions based on those follow-up questions. It's identifying the possible conditions and based on that, it is booking a doctor's appointment on your behalf. So fast tracking your ability to get specialized care.

This, what I just described, can be accomplished by chaining together multiple AI agents to form an agent tech AI workflow. And as suse, we are providing our customers the tools to build these workflows and also the implementation guides to get faster time to value and deploy them very fast. So that's another area that we are investing in.

And we have a couple more if you're interested. I, oh, We're interested, but I wanna stop here for a second. 'cause I want to dig into this.

Yeah. So this is like a pet peeve of mine. That's a great use case for an agent.

And what I like about it is it's a relatively narrow, you know, this is not looking to be your executive assistant. It does a specific job, almost like a robot, right? It's a single, almost a single task.

Take the information, determine write doctor, make an appointment. You as an individual, okay? You go, you have that at your doctor.

You have one of those at your dentist, maybe one of those at your barber. Uh, you know, before you know it, you may interact with two dozen agents like that. And then you may start saying, wait a second, I don't want to interact.

It's wasting my time interacting. I want my agent to interact with those agents. Mm-hmm.

Well, quite literally, pretty soon you're drowning in this a sea of agents, right? You have more agents. How are we going to, I mean, do we make a Kubernetes an orchestrator for agents?

Do we have a, a master agent that manages all of these agents? Are the agents ephemeral? Yeah.

Right? I used them that one time and they, they go away or whatever. Yeah.

The way I see it Thought of, yeah. I mean, that's, it's, it's rapidly evolving and it's a great question. There's a lot to think about there, but the way I see it that in terms of application of agents where they will be pervasive and ubiquitous is in internal organizational processes, which are automation of tasks that happen internally in an organization.

Mm-hmm. Like, whether it could be in a, in a scenario of software development or in an, uh, operational setup for any organization. That is where they'll, it'll be really powerful to provide those experiences.

And yes, there will be possibly a time where you'll have to orchestrate multiple agents. It's already happening, and the technology will evolve to manage all of that like it has over the last previous years. So, I mean, we, we have to see how it develops, to be honest.

It, it's going, it's going to be interesting. I, I, I think to, for all this to happen, you know, and, and quite frankly, look, there's a lot of people who are gonna watch this who say, well, that's my job. Right?

I was the one making the appointment. I was the one interviewing the patient and, and making, you know, not, not diagnoses per se, but preliminary, let's call it, um, you know, if you lift into, uh, Zuckerberg, they're gonna replace mid-level engineers with, with ai, uh, Benny, often Salesforce with their agent force replace a lot of sort of entry level lower level jobs. What do you say to people out there who are worried about these agents, these ais perhaps taking their jobs?

First of all, I completely understand. You know, a lot of people think that way. I think the more you get comfortable and familiar with any new technology, with any new technology, op initial adoption is always a challenge.

Mm-hmm. And when people start getting familiar with that technology and start using that and seeing the value it provides to their day-to-day lives, I think that's when folks realize the impact, the transformational impact, the gains in productivity, the operational efficiencies that they as individuals feel that organizations feel. So I understand that.

However, I feel that this is a curve for any new technology. This is a phase that every new technology, every new innovation goes Through. And we've all been through such phases across the last several decades.

Yeah. Whether it's The, the internet itself or the Cloud, the cloud, it did all, you know, And every single one of those made more jobs jobs than it took. It Did.

Exactly. So, I, I, I have a, a philosophy with this, and I, I tell it to a lot of young people who, you know, ask me about what kind of job should they go into? What do you think, you know, is a, is a worthwhile career?

And, and it's this, no one of us, you, me, or anyone, is gonna stop progress, progress stops for no man, as they say Well at it. Yeah. It's going to happen.

Yes. You've got two choices. You could sort of, you know, curl up in a fetal position and say, woe was me.

Or you could say, how can I leverage that to be the best me I could be? Yeah, That's Great. And that, that's life.

That's life. Right. That's a life lesson.

Exactly. Not just for ai, for any. And so my advice to people is always leverage it to be the best you can be.

Not don't, don't be afraid of it. Don't stick your head in the sand. Don't, you know, say, oh, it, it won't happen.

It won't happen. I cannot disagree with that. I completely agree for that.

Yeah. Okay. I think some of the jobs that are gonna come out of this will be, I mean, think about it, designing agents, but then, yeah, I, I did a, uh, I do a u uh, a LinkedIn live show every week, but a month ago, as part of this LinkedIn live, we were talking about this very subject, and I, I read somewhere that Mark Cuban mm-hmm.

Said that the world's first trillionaire won't be, uh, Elon Musk or Jeff Bezos or, or Zuckerberg or Bill Gates, or any of these, you know, billionaire people. The world's first trillionaire is gonna be someone who figures out how to use AI in a way we haven't thought of yet. Yes.

And makes that killer app. Yeah, for sure. 2 million views.

Wow. That's great. Crazy.

That's Great. Yeah. Um, So you see the interest, the Interest Is the interest of people wanting to learn Because they see, they see the opportunity.

Opportunity. You bet. Yep.

Yes. Absolutely. They're coming back to Zeka.

This is so Interesting to Zeka. We, I, we, we went, this is such, yeah. Yeah.

We went down that royal. Yeah. But we're at Han.

I mean, you, uh, you keynote it here today. Indeed. Tell us about your keynote.

Yeah, it was a great opportunity. I mean, uh, I shared something that I shared in the keynote. Uh, so when I got, you know, when I got asked to do this keynote, I was thrilled.

I was like, wow, what an opportunity. Let's do that. And, uh, then I started thinking, okay, so what do I do to, you know, capture some attention quickly, like in the first minute?

What can I do? How can I make this somewhat funny, somewhat witty, you know? And then I thought, okay, that shouldn't be a problem.

Uh, I'll just fire up my AI assistant. Yeah. Do some smart prompting, you know, come up with some interesting, funny anecdotes and jokes.

Oh, that should be pretty easy. Well, spoiler alert, it wasn't easy. So I must tell you that, Are we, are we blaming you for the prompt or the ai?

Oh, I think you, you must blame. It was you. So I, I, I genuinely believe so.

I, I stayed at a blank page for longer than I care to admit. Mm-hmm. So I, I did all of that, but I genuinely believe, I truly transformational.

But what I'm pretty sure it cannot do is turn me into a comedic genius. So that's, that's not gonna, Oh, have our limitations in us. It's not Gonna happen.

Right. So, yeah, I keynote back to the keynote. It was super fun, great opportunity, you know, uh, got an opportunity to share what SUSE is doing.

I shared two of those things already with you at the start of this, uh, right conversation. And let me share a couple of other things, right? So we are talking about AI assistance.

So when you, in any organization, when you deploy an AI assistant, how do you ensure that the output of that assistant is accurate? You know, so how can you ensure that, how can you ensure that it is not giving you an answer that is biased or that is violating your company's brand guidelines or your corporate policy, or spitting out confidential sensitive data for your organization? That's a real challenge.

The answer to that is something called guardrails. Guardrails as the technology, or as the name suggests, prevents any unintended behavior on part of these applications. So, from SUSE's perspective, what we are doing is that we are giving customers an opportunity to integrate guardrails into their AI environments by providing them the blueprints to do that with suse ai, number one.

Number two, you talked about partnerships as well, and we also announced a partnership with Infosys, where we are integrating Infosys responsible AI guardrails with suse ai, another way in which we can help our customers implement trustworthy, safe, ethical, and responsible ai. So that's another area. Absolutely.

We, we actually had some discussions this week with the, uh, Infosys partnership that it's very exciting. Also, the Reach Infosys is reach into the enterprise world worldwide and is, you know, probably one of the top four, I guess the guys, um, you know, a lot of people say, okay, so when's this all get real? Yeah.

So that's, that's a great question. Um, let me share. So over the last couple of years, let's say since November, 2022, since the, you know, advent of chat g bt so the world in general changed, you know, changed, let's say.

And, you know, a lot of organizations have invested a lot of money in, uh, driving POC environments with their AI implementation. Yes. And in talking to customers and partners over the last several months, we've realized that a lot of those POCs are not moving as rapidly into production deployments.

I agree with You. And one of the key challenges that we have seen is that as ai, small scale applications move from these POC environments to large production setups, costs start piling up to a point that they become Prohibitive. Prohibitive.

And that's a real challenge. Yeah. This, so what organizations must ensure is that their AI application or their language model, it's performing at its peak, right?

That they compute resources. The hardware, I Was just gonna say, Is fully utilized. They must leverage their investments fully.

And I add to that, because I wanna do a SUSE plug as well here while we're at it. So before you know, before you solve anything, you must ensure that you have compute visibility into the health and status of your AI infrastructure. Agreed.

And By the way, I might add, not just ai, I, I saw a recent survey, Kubernetes installs mm-hmm. Average, average use rate of your Kubernetes cluster, A CPU loads. Okay.

10%. Mm-hmm. 10%.

I'm not saying you gotta use 70%, no one's gonna use that. Mm-hmm. 10%.

Now you do get bursts. Yep. But average is 10%.

Wow. So then there goes, right. So you need visibility into the status, into the utilization of your infrastructure, because you cannot fix what you cannot see.

It's as simple as that. Yep. From our perspective with suse ai, we had today announced the launch of SUSE observability for ai.

Okay. So we are actually now giving our customers complete visibility into their AI infrastructure, into their relevant performance, token usage, latency drift utilization of their expensive GPUs into the response time of their applications, customers have's, control, And visibil. It's almost finops for ai.

Yeah. I mean, we are providing them the capability to understand which users and which applications are driving their costs. Is That available right now?

That's available starting April. We announced it today. Where can people get more information on all of the suse ai?

Yeah. com, check out the suse AI solution, and then reach out to a representative from suse via the website. That's what you get.

If you're already a SUSE customer, talk to your account executive, and that individual will be able to help you with more information on the same Abio. Thank you, Man. Thank you so much.

I, it was a great conversation. I Appreciate it. Hey, what a great way to end our SUSE con, uh, videos here with ai, the thing that's on everyone's minds.

We're gonna wrap, that's a wrap here from Orlando. Hey, I encourage you to go check out all of the videos we've done this week, as well as our coverage, as well as I think our fu Futura research analysts. Mitch Ashley was here, and I think he has a note on this.

So a lot of Scon coverage, Firmex Strong and Futurum, and we'll be covering this whole area, of course, going forward in just a few weeks. We're in London for Q Con. That's amazing.

And of course, s yeah. Next Well absolutely. And we'll be continuing the ation.

Until then, this is Alan Shimmel for Techstrong at Han. We're out.