In this episode, our Andy Leonard and Frank La Vigne sit down with Ronen Dar, the co-founder and CTO of Run AI, to explore the world of artificial intelligence and GPU orchestration for machine learning models.

Ronen shares insights into the challenges of utilizing GPUs in AI research and how Run AI's platform addresses these issues by optimizing GPU usage and providing tools for easier and faster model training and deployment. The conversation delves into the concept of fractional GPU usage, allowing multiple workloads to run on a single GPU, making expensive GPUs more accessible and cost-effective for organizations.

Links

• Run AI https://www.run.ai/
• Acquired Podcast Episode on Nvidia https://www.acquired.fm/episodes/nvidia-the-machine-learning-company-2006-2022

Show Notes

04:40 GPU technology enabled for cloud AI workloads.

07:00 RunAI enables sharing expensive GPU resources for all.

11:59 As enterprise AI matures, organizations become more savvy.

15:35 Deep learning, GPUs for speed, CPUs backup.

16:54 LLMs running on GPU's, exploding in market.

23:29 NVIDIA created CUDA to simplify GPU use.

26:21 NVIDIA's success lies in accessible technology.

28:25 Solve GPU hugging with quotas and sharing.

31:15 Team lead manages GPU quotas for researchers.

35:51 Rapid changes in business and innovation.

40:34 Passionate problem-solver with diverse tech background.

43:38 Thanks for tuning in, subscribe and review.

Greetings, listeners. Welcome back to the Data

Driven Podcast. I'm Bailey, your AI host with

the most data, that is, bringing you insights from the ether

with my signature wit. In today's episode, we're

diving deep into the heart of artificial intelligence's engine room,

GPU orchestration. It's the unsung hero

of AI research, optimizing the raw power needed to fuel

today's most advanced machine learning models. And

who better to guide us through this labyrinth of computational complexity than

Ronan Darr, the cofounder and CTO of Run AI, the

company that's making GPU resources work smarter, not

harder. Now onto the show.

Hello, and welcome to Data Driven, the podcast where we For the emergent fields

of artificial intelligence, data engineering, and overall data

science and analytics. With me as always is my favoritest

Data engineer in the world, Andy Leonard. How's it going, Andy? It's

going well, Frank. How are you? I'm doing great. I'm doing great. It's been,

we're We're recording this February 1, 2024. And as I said to my

kids yesterday, January has been a long year.

We're only, like, 1 month into the year, and it was it was a pretty

wild ride. But I can tell we're gonna have a blast today,

because we're gonna geek out on something that I kinda sort of understand,

but not entirely, and it's GPUs. And in the virtual green room, were chit

chatting with some folks, and, but let me do the formal introduction

here. Today with us, we have doctor Ronadhar, cofounder and CTO

of Run AI, A company at the forefront of GPU

orchestration, and he has a distinguished career in technology.

His experience includes significant roles at Apple. Yes, That

apple. Bell Labs. Yes. That Bell Labs.

And at Run AI, Ronan is instrumental in optimizing

GPU usage For AI model training and deployment,

leveraging his deep passion for both academia and startups.

And, Run AI is a key player in the, and he is a he

and Run AI are key player in the AI revolution. Ronan's

contribute Contributions are pivotable in shaping and powering the

future of artificial intelligence. Now I will add that in

my day job at Red Hat, Run AI has come up a couple of times.

So this is definitely, definitely

an honor to have you on on on the show, sir. Welcome.

Thank you, Frank. Thank you for inviting me. Hey, Andy. Good to

be here. I love it. Love Reddit. We're a big

fan of Reddit. We're working closely with many people in

Reddit, and love that. Right? Love OpenShift,

love Reddit, love Linux. Yeah. Cool. Cool.

Yeah. So so for those who don't know exactly, I kinda know

what, your Run AI does, but can you explain exactly

What it is run AI does and why GPU

orchestration is important. Yes.

Okay.

So run AI is, software,

AI infrastructure platform. So we

help machine learning teams to get much more

out of their GPUs, And we provide

those teams with abstraction layers and tools

so they can train models And deploy models

much easier, much faster. And

so We started in 2018, 6 years

ago. It's me and my cofounder, Omuri. Omuri is the CEO.

He's, he's amazing. I love him. We We know each other for many

years. We we met in the academia, like, more than 10 years ago,

and and we started running AI together, and We started

running AI because we saw that there are big challenges

around, GPU's, around orchestrating

GPU's and utilizing GPU's. We saw back then

in 2018, the GPUs are going to be very very important.

It's like the basic a a component in

that any AI company need to train models,

right, and deploy models. So we saw that GPUs are going to be critical, but

there are also a lot of challenges with, with utilizing GPUs.

I think back then, GPUs were relatively new In

the data center, in in the cloud.

GPU's were very known in the gaming

industry. Right? We spoke before on gaming. Right? Like, a lot of

key things there that GPU's has has has been enabled

enabling, But in the data center, they were relatively new and the

entire software stack that is that

is running the Cloud in data center As was built for

traditional microservices applications that are running

on commodity CPUs And AI workloads are different, they are

much more compute intensive, they they

run on on GPUs, maybe on multiple nodes of Meet to point

machines of GPU's, and GPU's are also very different.

Right? They are expensive, very scarce in the data center.

So The entire software stack was a bit for something else

and when it comes to GPUs, it was really hard for many people to to

actually manage those GPUs. So we came in And, and we

saw those gaps. We've built run AI on top of

cloud native technologies like Kubernetes and containers. We're

big fans of Of those, technologies, and

we added components around scheduling, around

the GPU fractioning. So we enable

multiple workloads to run on a on a single GPU and

essentially all the provision GPU's. So we build this Engine which we

call cluster engine that runs in in in GPU

clusters. Right? We help machine learning teach to pull all of their GPU's into

1 cluster, Running that engine, and that engine provides a lot of

performance and lot of capabilities from those GPUs. And

on top of that, we built this control plane And

and tools and for machine learning,

teams to run the Jupyter Notebooks, to run

training jobs, batch jobs to deploy their models, right, to just to to

have tools for the entire life cycle of AI

from Training models in the lab to taking those models into

production and running them and serving actual users.

And That's the platform that we've built, and we're working with machine

learning teams across the globe and on just managing,

orchestrating, and letting them Get much more out of their GPUs and essentially

run faster, train more than faster and in much easier way and

deploy those modules In a much easier and faster and more efficient

way. Yeah. The thing that blew me away when I first heard of Run

AI, and this would have been, 2021

ish. No. 20 early

2021, I would say, And, it was the

idea of fractional GPU's. Right? So you can have 1,

I say 1, but, know, it's realistically, it's gonna be on, but you you can

kind of share it out, which I think and we were talking in the virtual

green room about how, you know, some of these GPU's,

If you can get them because there's a multi month, sometimes multi

year supply chain issue. I mean, these things are expensive bits of

hardware, and I think the real value, correct

me if I'm wrong, is, like, well, you know, if you I was talking to

somebody the other day, and and we're basically talking about how we can,

you know, if you get if you get, like, 1 laptop with a killer

GPU, right, that GPU is really only useful to that 1

user, Whereas if you can kind of put it in a in a in a

server and use something like RunAI, now everybody in the organization can do

that. And these are not trivial expenses. I mean, these are like, You know,

you sell a kidney type of costs here.

Yeah. Absolutely. So Absolutely. First of all, GPUs

are expensive. They cost a lot. Right?

And we provide, Technologies like fractional GPUs and

other technologies around scheduling that allows

teams to share GPUs. Right. So we used book on

GPU fractioning. So that's 1 one day of

sharing where you have 1 GPU, which is really expensive.

And Not all of the workloads are

AI workloads are really compute intensive and require the

entire GPU or, you know, maybe multiple GPUs. There are

workloads like Jupyter Notebooks where you have

researchers that just

Debugging their code or cleaning their data or doing some simple stuff,

and they need just fractions of GPUs.

In that case, if you have, a lot of data scientists,

maybe you wanna host all of their notebooks On

a much smaller number of GPUs because, right, each

one of them, it's just fractions of GPUs. Another big use case

for fractions Of GPUs is inference.

So now all of the models are huge

and And doesn't fit into, the memory of 1

GPU, and in computer vision,

there are a lot of Models that are relatively small,

they run on GPU, and you can essentially host multiple of

them on the same GPU. Right. So you can have instead of

just 1 computer vision model running on GPU, host 10

of those models on the same GPU and get Factors of

10 x in, in your cost, in your,

overall throughput of, of inference. So that's That's one

use case for fractional GPU, and we're investing heavily just

building that technology. Another layer

of sharing GPUs Comes where you

have maybe in your organization multiple teams

or multiple projects running in parallel. So

for example, may open AI, they now are working

on gpt5. It's 1 project. That project needs a

lot of GPUs And they have more projects. Right?

More research project around alignment or around,

reinforcement learning. You know? DALL

E. Like, they they they have more than just 1 project. Then DALL E and

they have multiple models. Right? Exactly. They have. Right? So each

project needs Needs GPUs. Right? Needs a lot of

GPUs. So if you can instead of

allocating GPUs Entirely for each project,

you could essentially pull all of those GPU's and share

them between the those different projects, different teams,

And in times where 1 project is idle and not

using their GPUs, other projects, other teams can share

can get access to those GPUs. Now orchestrating all of

that, orchestrating that sharing of resources between

projects, between teams can be really complex And

requires this advanced scheduling, which

which we're bringing into the game. We're bringing

those scheduling capabilities from the high performance computing world

known on those schedulers. And so we're bringing Capabilities

from that world into the cloud native Kubernetes

world. Scheduling around batch batch scheduling

fairness, Algorithms, things like that, so teams and projects

can just share GPUs in a simple and efficient

way. So those

are the 2 layers of sharing GPU's. Interesting. And and

I think that I think as As this field matures

and it matures in the enterprise, I think you're gonna see organizations

kind of be more,

more more more I think savvy about, like, okay, like you said, like, data scientists,

if they're just doing, like, you know, Traditional statistical modeling really doesn't benefit

from GPUs, or they're just doing data cleansing, data engineering.

Right? They're probably gonna say, like, well, Let's run it on this cluster, and

then we'll break it apart into discrete parts where, you

know, then we will need a GPU. And I also like the idea that, you

know, you're you're basically doing What what I learned in college,

which was time slicing. Right? Sounds like this is kind of, like, everything old is

new again. Right? I mean, this is, Obviously, you know, when you're when you're

taking kind of that old mainframe concept and applying it to something like Kubernetes,

orchestration is gonna be a big deal, because these are not systems that were Not

built from the ground up to have time slicing. Is that a is that a

good kind of explanation? Yeah. Absolutely.

Absolutely. I like I like that analogy. Yeah. Exactly. Time

slicing it's, it's 1 so

1 implementation, Yeah. And that we

enable around fractionalizing GPU's,

and I agree when you have resources, It

can be different kind of resources. Right? It can be CPU

resources and networking were also,

You know, as people created that technology to share the

networking and communication going through those networking, but just the

bandwidth of the networking. We're doing it

for GPU's. Right. Sharing those

resources. And I think now it interestingly,

LLMs I also becoming a kind

of, resources as well, right, that people need access

to. Right? You have those models, you have GPT, JGPT.

A lot of people are trying to get access to

that resource, essentially. And I think it's interesting,

because you kinda pointed this out, but it it it's something that I think that

if you're in the gen AI space, you kinda don't it's so it's obvious

like error. You don't think about it. Right? But when when you

get inference on traditional, I somebody once referred to it

as legacy AI. Right. But where

the infrared side of the equation, you don't really need a lot of compute power.

Right? Like, it's not really a heavy lift. Right? But with generative

AI, you do need a lot of compute on

I I guess it's not really inference, but on the other side of the use

while it's actually in use, not just the training. Right. So traditionally,

GPU heavy use in training, and then inference, not so

much. Now we need heavy use before, after, and during,

which I imagine your technology would help because, I mean, look, I love chat I

love chat g p t. I'm one of the 1st people to sign up for

a subscription, But even, you know, they had trouble keeping

up, and they have a lot of money, a lot of power, a lot of

influence. So I mean, this is something that if you're just a

regular old enterprise, this is probably something they struggle

with. Right? Right. Yeah. I absolutely

agree. It's like amazing point, Frank.

So 1 year

ago, the inference use case on

GPU's. Wasn't that big. Totally agree. That's also what we

saw in the market.

Deep learning Convolution neural networks were

running on GPUs,

mostly for computer vision applications,

But they could also run on CPUs and you could get,

like, relatively okay performance.

If you needed maybe, like, a very low latency, then

you might use GPUs because they're much faster and you get much

lower latency. But

it was, it was all, and it's still very

difficult to deploy more than it's on GPU's Compared to just deploying

those models on CPUs, because deploying more than deploying applications on

CPUs, you know, people are doing for so many years.

So

many times it was much easier for people to just deploy their

models on CPU's And not on GPUs, so that was, like, the

fallback to CPUs. But

then came, and as you said, chair GPT was introduced, A

little bit more than a year ago, and that generative

AI use case just blown. It was blown. Right? And it's

it's inference essentially. And those models are

so big that they can't really run on

CPU. They, they LLMs are running in production on

GPU's and now the inference use case on

GPU's is just exploding In the market

right now, it's really big. Is a lot of demand for

GPU's for inference And

if for open AI, they need to support this

huge scale that I guess, just

Just them are seeing such scale, maybe a little, a

few more companies, but that's like huge, huge scale.

But I think that we will see more and more companies

building products based on AI, on

LLMs, And we'll see more and more

applications using AI, which

then that AI runs on on GPU. So That is going to go

and that's the that's an amazing new market for us around

AI and for me as a CTO, it was so fun to

Get into that market because it now comes with

new problems, new challenges,

new use cases Compared to deep learning

on on GPS. New new pains because

the models are so big. Right? Right. And

challenges around cold start problems, about auto scaling,

about, About

just, giving access to LLMs. So a lot of

challenges, new challenges there. We at Tron AI will studying those problems

and we're Now building solutions for those problems,

and I'm really, really excited about the Inference use case. That

is very cool. So just, going back a little bit.

I was trying to keep up. I promise. But Run AI is

I I get Run AI Run AI's platform

Support fractional, GPU usage.

It it also sounds to me, maybe I misunderstood,

That in order to achieve that, you first had to or

or maybe along with that, you made it possible to use multiple

GPUs. You've you've created Something like

an API that allows, companies

to take advantage of multiple GPUs or fractions of

GPUs. Did I Did I miss that? No, that's

right. That's right, Andy. And Okay.

So we've built this, way of,

For people to scale their workloads from fractions

of GPUs to multiple GPUs within 1 machine,

Okay. To multiple, machines. Right? You

have big workloads running on on multiple nodes

of GPUs. So Think about it when you have

multiple users each running their own

workload. Some are running on fractions of GPUs. Some are

running batch jobs on on a lot of

GPUs. Some Deploying models and running them on

in inference, and some just launching their Jupyter

Notebooks. All of that is happening on the same

pool of GPU's, same cluster. So you need

this lay of orchestration of scheduling just to

Manage everything and make sure that everything getting there

right, access the right, and and

and g p u's And everything is scheduled according to

priorities. Yeah. Well, being just, you know, a

mere data engineer, Here talking about all of that

analytics workload. That that sounds very

complex. So and as you

mentioned earlier, you know, you were talking about how traditional coding

is targeting CPUs, and that's my background.

You know, I've written applications and and done data work targeted for

traditional work. I can't imagine, just how complex

that is, because GPUs came into AI

as a unique solution,

designed to solve problems That they weren't really built

for. You know, GPUs were built for graphics, and you didn't

manage that. But the fact that They have to be

so parallel, internally. I think just added this

dimension to it. And I don't know who came up

with that idea, you know, who thought of, well, goodness, we could we could

use all of this, you know, massive parallel processing to To

to run these other class of problems. So pretty

cool pretty cool idea, but I just I yeah. I'm amazed at even

cooler than that. Because Yeah. Yeah. A wise man once told me,

he goes, GPU's are really good at solving linear

algebra problems, And if you're clever enough, you can

turn anything into a linear algebra problem.

And even simulating quantum computers when I was kind of, like, going through that,

I was like Mhmm. You know, like, gee, looks like looks like this

will be useful there too. Right? Like so it's an it's an interesting,

It's an interesting thing. So, like, you know, everyone is, you know,

everyone's talking about how this is, you know, we're in the hype cycle, but I

think if you're in the GPU space, you have Pretty good run because one,

these things are gonna these things are gonna be important. Right? Whether or not, you

know, hype cycle will will kinda crash, and how what that'll look like.

Think they're gonna be important anyway. Right? Because they're gonna be just the cost of

doing business, table stakes, as the cool kids like to say. But

also, over the next horizon, Simulating quantum

computers is going to be the next big hype cycle.

Right? Or one of them. Right? So like it's

it's it's a It's a foundational technology. I think that we

didn't think would be a foundational technology even like 6 7 years

ago. Right? Yeah.

I go with a few things that you said.

Regarding the Parallel computation, right? And just running

linear algebra calculations on GPU's

and accelerating such workloads.

In Nvidia, I love Nvidia, Nvidia

has this big vision, and they had big

vision Around GPU's already in 26 when

they built CUDA. Yep. Right. So

They've been good at just for that. Right? The GPU's were

used for graphics processing, For gaming.

Right? Great use case. Great market.

But they had this vision of bringing more

Applications to GPU is just accelerating more applications

and mainly applications with a lot of Linear

algebra calculations. And they

created that, they created CUDA

To simplify that. Right? To allow more

developers to use GPUs because just using GPUs

directly, that's so complex. That's so hub.

So we've built CUDA to bring more developers, to bring more

applications and they started in 20

2006, but think about the

big breakthrough in AI, it happened just in

2012, 2013 with

AlexNet and the Toronto researchers

who used G2 GPU's actually, because they

trained Alex Net on 2 GPU's and they had

CUDA, so for them it was feasible To train their

model on a GPU. And that was the new thing that they did.

They were able to Train much bigger model with

more parameters than ever before because they use

GPU's because the training Process ran much

faster. And,

and, and that triggered the entire

revolution, the Die hyper on the AI that we're seeing now. So

from 26, when Nvidia started to build CUDA until

2013, right, 7 years, Then we started to see

those big breakthrough. And in the last decade,

it's just exploding, and we're Seeing more and more applications.

The entire AI ecosystem is running on on an

on GPUs. So that's amazing to see. It's impressive.

And, like, People don't realize, like, the the revolution we're seeing today

really started in 2006, like you said. I didn't even put the 2 and 2

together until I was listening to a podcast. I think it's called Acquired,

And really good podcast. Right? Like, I they don't pay me to say that or

whatever, but they did a 3 hour deep dive on the history of

NVIDIA. 3 hours. I couldn't stop listening.

Right? Like Nice. You know Yeah. We tried a long form, like, multi hour

podcast. We Weren't that entertaining, apparently. But the way they

go through the history of this where it was basically Jensen Huang. Hopefully, I said

his name right. He was, like, we wanna be a player, not just in gaming,

but also in scientific computing. This is 2005, 2006,

which at the time seemed kind of, like, Little out there, little kooky.

But what you're seeing today is, like, the the fruits and the tree the the

seeds that he planted, I, you know, almost 20 years ago, like, 19,

20 years ago. So, you know, it's you know, when people look at

NVIDIA and say it's overnight Success. I'm like, well, I don't know about that, but,

you know, but no. I mean, you're right. Like, you know and it's

probably not a coincidence that once they made it easy to take these

Multi parallel processor. Say that 10 times

fast on a Thursday morning. But also

make it so it's a lot easier for developers to use. Right? And I'll quote

the great Steve Ballmer, developers, developers, developers. Right?

So, it's it's, it's just fascinating, like and

and I think that, you know, we've really on Leafy a

gate of creativity in terms of researchers and applied,

research, and, I mean and I think that what's really cool

about your Product is that you're you're kind of making this what is

now a sparks resource, maybe in some fashion

of time, GPU's won't Cost an arm and a leg.

But, like, for now, I think I think the one thing that I've seen

that I think is, not obvious For the casual

observer is if you can if an

organization, like a large enterprise, can pull their resources, they have a lot more

money to buy better GPUs, And you offer a platform where

everybody can get a stake in it. Right? As opposed to, you know you know,

that department is gonna hog everything. Right? You know, you and and and and,

here's a question. Do you do you have, like, an audit trail where you could

kinda, you know, figure out, like, you know, Andy's department's really

hogging the GPUs. No. No. No. It's Frank. Frank is like mining Bitcoin or

whatever. Like, do you do you have some kind of, audit trail like that?

Yeah. I I love that you mentioned hugging, We

GPU hugging. We Mhmm. We use that term as well.

Right? Because it it's so difficult sometimes to get

access to GPUs. So when you get access to GPU

as a researcher, as a member practitioner,

you don't wanna Let it go. Right. Cause if

you let it go, someone else would take it and hug it. Right.

So you're getting this GPU hugging problem.

What we do to solve that is

that we do provide monitoring and visibility

tools into who is using what, and who is actually

utilizing their GPU's, and so on, but more

than that We

allow the researchers just to give up their GPS and not hardware

GPS because we provide this, Concept of

guaranteed quotas. So each researcher or

each project or each team has their own guaranteed

quotas of GPU's That are always available for them

whenever they will get access to the the cluster, they will get like, you

know, the the 2 GPUs or 4 All the quarter of

GPU's it's guaranteed. So they can

just let go their GPU's and not hug them. That's one

thing. The second thing is that they

can also go above their quota. They can

use the GPUs of Other teams or other users, if

they are idle, and they can run this preemptible jobs

in an opportunistic way, utilize those GPUs.

And so in that way, they are not limited

to fixed quotas, to help limit

quotas. They can just take as many GPUs

as they want from their clusters if those GPUs are available

in idle right but if someone will need those gpus

because those gpus are guaranteed to them we will make sure our

scheduler The Run AI schedule that the Run AI platform will make

sure to preempt workload

and give those Guarantee GPUs to the right users.

Oh, that's cool. Alright. So 1 last

question before we switch over to the the stock questions, cause I could geek

out and look at this for hours. Yep. This could be a

long form. Sure. This could be. Yeah. And that's and I I wanna be respectful

of your time because you're an important guy, and it's also late where you are.

So who deals with this? Like, who would set up these quotas? Is it

the is it the is it the data scientist? Is it IT ops? Like, who

do you obviously, the data scientists, Researchers, they all

benefit from this product. But who's actually administering it? Right? Like,

who is it you know, do I have to talk to, you know,

Say pretend Andy's in ops. Do I have to say, hey, Andy. I really need

a boost in my quota. You know, like, I mean, who does it? Or do

or my this sounds like you as I say it, I'm like, yeah, that wouldn't

work. Like, I'm the researcher. I'm gonna turn the dial up on my own. Like

like, who's who's who's the primary? Obviously, we know who the prime

primary beneficiary is, but who's the primary user?

So okay. Great. So if you have a team, right, if if

you're a team of researchers, all all of you Need access to

GPU, so maybe the team lead

is the one who's managing the quotas for the different

team members. And if you have multiple teams,

then you might have a department manager or an admin of the

cluster or platform owner that will Allocate the

quotas for each team, right? And then those teams would

manage their own quotas within That's what

they they they were giving. Right? So it's like a a hierarchical

thing in a hierarchy manner. People can manage their own

quota, their own, priorities, their own access to the

GPUs within their teams. Okay.

So it's kind of like a hybrid of, like, you know, it's like a budget

almost. Right? Like, you know, you get this much, Figure it out

about yourselves. Exactly. So we're trying to decentralize

the how the quotas are being managed and how the GPUs are being accessed.

So, you know, I'm giving as much power, as much

control to the end users as possible. Sure. That's

It sounds like a great administrative question, very

important. And I imagine, because a little bird told

me that you're not the only, you know, your your

provisioning provisioning of these GPU resources

is not the only thing that, enterprises have to deal

with. So it's an it's an interesting just GPUs.

It's compute. Like, it's not a Sure. It's not it's not limited. Although, because

of what you said, you know, Managing GPUs is an order of magnitude harder

because they were never really built for this. Right? Like, this kind of Right. You

know, we're talking about technology that wasn't really in the server room until Few

years ago. Right? This isn't a tried and true kind of this is

how it works, you know? Right. But we hit that point in the

show where we'll, switch the preform questions.

These are not complicated. I mean, you know, we're not we're not Mike

Wallace or, like, you know, 60 minutes or whatever. We're not trying to trap you

or anything. But since I've been gabbing on most of the show, I

figured I'll get Andy kick this off. Well, thanks, Frank. And I don't think

you were gabbing on. You know more about this So now I do. So I'm

just a lowly data engineer. I'll plug No. You if you

will. Data engineers are the heroes we need. Well

well, I'm gonna plug Frank's Roadies versus Rockstar's,

writing on LinkedIn. It's it's good articles about this.

But, let's see. How did you,

how did you find your way in into this field?

And, did did this feel fine you or did you find it?

This feel totally fine found me. Awesome.

Yeah. I I've

I did my post doc, and I've been in Bailabs.

And Jan Hakon came to Bell Labs and

gave a presentation about AI. It was around 2017,

And Jan Hakun spent a lot of years in Bell Labs,

and his presentation was amazing. And

When I heard him talking about AI,

I I said, okay, that's the space where I wanna be. It's going to change

the world. There is this New amazing technology here that

is going to change everything. And I knew that I want to start

a company In the AI space for sure.

Cool. That's a good answer. So cool.

Yeah. That's cool. I was at Bell Labs,

doing a presentation a while ago, and somebody I didn't realize that he

worked at Bell Labs because, like, you know, the guy was like, no. No.

He used to work here, like, in this building. I was like, no way. Because

I knew him as the guy from NYU. Right? Like, that's who I thought. Right.

For the guy from from Meta. Yeah. And now the guy from Meta. Right? Like

so it's interesting how that how that you know? They have

this amazing pictures from the nineties where they

run like deep learning models on very old pieces

and, And recognizing like,

numbers on the computer. Maybe you saw those pictures like amazing

Emmis. It's the Emmis problem. Is that Yep.

Right. Exactly. Exactly. Cool.

So second question is, what's your favorite part of your current job?

That everything is changing so fast.

Things are moving so fast right away in this business for 6

years, and the entire

space is moving and

advancing. And so many people are working in

this field A new innovation, new tools,

new new advancements are are getting out every day.

You know, just 6 years ago, it was about deep learning and computer

vision. And now it's about language models

And generative AI, and we're gonna just at the start,

right, there are so many amazing things that are going to happen

in this space, and I love it. Absolutely.

So we have 3 fill in the blank

of sentences here. The first Is complete this

sentence when I'm not working, I enjoy blank.

You'll get a you'll get a very boring And

so this is just spending time with

friends and family, because I think

That I'm always working. It's like, if you ask my wife,

she'll tell you that I'm working 24 hours. And

Yeah. So I don't have much time that I'm not working

in. So when I I do I'm not when I'm

not working then I'm trying Trying to be with my kids and my

wife and friends. Cool.

Cool. The 2nd complete the sentence. I think

the coolest thing about technology today is

blank. And this, I really wanna hear your perspective on that.

Yeah. I think everyone will say AI, right? Or something in

AI. Yeah.

I think there are so many

new innovations that are coming around LLMs.

I think everything relating to

searches, right? Searching in data, in getting

insights From data, it's all going to change. We're going to have

a new interface. Right? Just getting

insights from data from And natural with

natural language, oh, you know, no SQL and, you

know, needing to programming and stuff like that.

Just With natural inter language, you could

do amazing stuff with data. I think,

We're seeing this,

advancement in, And like digit

digital twins right now. You can,

you can, Fake my voice

and your voice and fake my image and your image. And,

and, and, you know, In in the

future, we'll have digital twins of us, right,

doing this stuff. That would be amazing. So a lot of

amazing stuff are going to happen in the next few years

for sure. Very cool. Our last complete sentence.

I look forward to the day when I can use technology to

blank.

To have a robot in my house.

Yeah. Yeah. You're swapping the flow in instead of

me doing that, right, cleaning dishes and things like that.

If that would happen, that would be amazing. Right? That's a that's a

good answer. Yeah. I I agree. I have I have 3

boys, 4 dogs. So, like, cleaning is safe.

Yeah. Yeah. I'm a heavy cleaning. Ranging from, like, 1 to, like,

a teenager. So it's it's, and and and fighting

with them to, Like, empty the dishwasher is takes a lot more mental

energy than it should, but that's probably a subject for another

type of show.

The next question is share something different about yourself,

and we always like to Joke like, well, let's just make sure that we keep

our clean Itunes rating. So Yeah. Yeah. What

what yeah. Well, I I This

is a hard question, I needed to think about it.

So, I found 2 answers that I can say. So one

is about my professional life, right, I think that

it's somewhat different that I'm coming this With back from

the academia and the industry. So I love academia. I love to research

problems. I love to understand problems in in a deep

way And combining it with startups in the industry.

And, and in my past, I worked for cheap companies, for hardware

companies. I work for Intel, for startup, and for Apple. I

did cheap stuff, and now 1 AI is a software company, so really

like a diverse background of Academia, hardware,

software, so I love that, and, like, I love to do

with few things, and so that I think is different.

And the 2nd answer that I could find

is, that I have a nickname that goes with me

since my high school days, Which is, the Duke.

The Duke. All of them all of them are calling me the Duke. It's like,

they don't call me Ronan, the the Duke. So That's funny.

Yeah. That's awesome.

Automotive is a sponsor of, Data Driven,

And you can go to the datadrivenbook.com.

And if you, if you do that, you can sign up for a free

month Of Audible. And if you decide later to

then join Audible, use one of their their sign up plans,

then Frank and I get to Split a cup of coffee, I think,

out of that. And, every little bit helps. So we really

appreciate that when you do. What we'd like to ask

Yes. Do you listen to audiobooks? And if you

do okay. Good. I see you nodding. So do you have a recommendation? Do you

have a favorite book or two you'd like To share. Yeah.

So I'm a heavy user of, audible. I'll give them

the, a classical book with Classical for

entrepreneurs, on their how the hard things

about how things from by Ben Horowitz,

it's Classic book, love it, really did a lot of impact

on me, I read it when we started run AI

And I recommend it for every

entrepreneur, to read it and for everyone to read it. It's like a

Cool. Amazing book. Yep. Awesome. I

have a flight to Vegas this next week, so I'll definitely be listening to

it then. And finally, where can people learn more about you

and run AI? And best

place will be on our website, Run dot a I.

Yeah. And on social. LinkedIn, Twitter, we'll

we'll do. Awesome any parting thoughts

I really enjoyed this episode love to speak about gpu's love the ai Based

on it, I had a lot of fun. Thank you for having me here. Awesome.

It it was an honor to have you, and every once in a while, Andy

and I will do deep dive kinda shows. We love to invite you back if

you wanna do 1 just on GPUs, because I know where my knowledge

drops off, you probably could pick up on

that. And with that, I'll let the nice

AI British lady end the show. And just like

that, dear listeners, We've come to the end of another enlightening

episode of the data driven podcast. It's always a

bittersweet moment like finishing the last biscuit in the tin,

satisfying, yet leaving you wanting just a bit more. A

colossal thank you to each and every one of you tuning in from across the

digital sphere. Without you, we're just a bunch of

ones and zeros floating in the ether. Your support is what

keeps this digital ship afloat, and believe me, It's much appreciated.

Now, if you found today's episode as engaging as a duel of wits with

a sophisticated AI, which I assure you, is quite

enthralling, then do consider subscribing to Data Driven.

It's just a click away and ensures you won't miss out on our future true

adventures in data and tech. And if you're feeling

particularly generous, why not leave us a 5 star review?

Just like a well programmed algorithm, your positive feedback helps

us reach more curious minds and keeps the quality content flowing.

It's the digital equivalent of a hearty handshake.

So, until next time, keep those neurons firing, those

subscriptions active and those reviews glowing. I'm

Bailey, your British AI lady, signing off with a heartfelt

cheerio and a reminder to stay data driven.