Rise of AI Conference 2022 · Berlin · Applied AI Stage

Solving the World’s Greatest Challenges

Jaap Zuiderveld · NVIDIA

At Rise of AI 2022 in Berlin, on the Applied AI Stage, Jaap Zuiderveld of NVIDIA argues that NVIDIA is best understood as a software company rather than a hardware maker: of its 24,000 employees, 22,000 work in engineering and 70 percent of those do software. He walks through NVIDIA's stack of pretrained models and SDKs, including Modulus, Clara, Riva, Maxine, Merlin, Omniverse Avatar, Drive, Isaac, Metropolis and Holoscan, and gives customer examples for each, from Siemens Energy's wind turbines to Amex's call-center voice recognition. He stresses that acceleration matters because waiting for results wastes the time of expensive data scientists, and cites growth figures: 3 million developers, more than 1,300 partners, 25,000 customers and 2,700 accelerated ISV applications. He highlights partnerships with Daimler on self-driving software and with BMW on an Omniverse digital twin of its factories, and describes AstraZeneca's use of an NVIDIA supercomputer to speed up COVID vaccine development. He closes by describing simulating the physical world in virtual environments as NVIDIA's next major wave.

Key takeaways

  • NVIDIA describes itself as an engineer-led software company: of 24,000 employees, 22,000 work in engineering and 70 percent of those do software.
  • NVIDIA offers pretrained models across domains, including Modulus for physics simulation, Clara/MONAI for healthcare imaging, Riva for speech, Maxine for video, and Merlin for recommendations.
  • NVIDIA partners with Daimler on self-driving software targeted for the road in 2024 and with BMW on an Omniverse digital twin spanning multiple factories.
  • NVIDIA counts about 3 million developers worldwide, more than half of whom joined in the past year, and accelerates over 2,700 ISV applications.
  • AstraZeneca used an NVIDIA-built supercomputer in Cambridge to accelerate COVID vaccine development by more than a year.

Detailed notes

NVIDIA as an engineer-led software company

Zuiderveld opens by saying most people know NVIDIA only from gaming and do not know the company otherwise, even though a lot of what is done in AI today comes from gaming technology. He states that NVIDIA has 24,000 employees, of whom 22,000 work in engineering, and that 70 percent of those do software engineering against 30 percent hardware engineering. On that basis he describes NVIDIA as an engineer-led software company rather than a sales or hardware company.

The pretrained model stack

He presents NVIDIA's full stack of pretrained models and SDKs, meant to make the technology easier to use and to "democratize" it. Examples he gives: Modulus, a pretrained physics-based simulation model, used by Siemens Energy to develop wind turbines; Clara/MONAI, a healthcare model for image recognition developed with King's College Hospital and the NHS in the UK, using federated learning to protect patient data; Riva, a natural language processing model used by Amex so call-center staff can better recognize customer voices and questions; Maxine, which enhances video images in tools like Teams or Zoom; Nemo, an advanced NLP model comparable to Alexa; Merlin, a recommendation engine that lets companies compete with players like Amazon without training a full model from scratch; Omniverse Avatar; Drive, developed with Daimler for self-driving cars planned for the road in 2024; Isaac for robotics, with Amazon using it to reduce accidents and improve routing in distribution centers; Metropolis for video surveillance, used by most major airports; and Holoscan for medical imaging and instruments, with Siemens as a major user.

Why acceleration matters

Zuiderveld argues that speed changes how businesses operate: getting an answer in five minutes instead of a day makes decision-making faster and more agile. He notes that the most expensive resource in a company is often its data scientists, so making them wait for results is costly regardless of hardware cost. He cites NVIDIA figures: an HPE server with one Hopper-architecture GPU is already faster than the fastest supercomputer from ten years ago; NVIDIA has 3 million developers worldwide, more than half of whom joined in the last year; it follows 10,000 startups worldwide, 450 of them in Germany; it accelerates more than 2,700 applications from independent software vendors; and it has more than 450 SDKs and models and 8 million CUDA developers. He adds that NVIDIA has about 3,000 people in Israel following its acquisition of Mellanox, and describes a growth rate of 50 to 60 percent year over year, including in Europe and China.

Industry examples and the next wave: digital twins

He gives further customer examples: AstraZeneca used an NVIDIA-built supercomputer in Cambridge to develop its COVID vaccine, accelerating the process by more than a year; the Human Brain Project uses NVIDIA technology to research the origins of Parkinson's disease; and Red Bull's F1 design work benefits from NVIDIA technology. He states NVIDIA has 25,000 customers globally and more than 1,300 partners. He names simulating the physical world in a virtual environment as the next major wave, citing that more than 8.6 billion compute hours are already used yearly for rendering in media and entertainment. As an example he shows a video of BMW building a complete digital twin of its factories on Omniverse to plan production, improve logistics and coordinate global teams virtually, expanding the initiative to three additional factories totaling more than six million square meters.

Q&A highlights

Asked how he represents NVIDIA in Germany, Zuiderveld says he works with a local team of about 1,000 people, including significant software engineering in Munich and Berlin, where NVIDIA's Omniverse development headquarters is located, plus a commercial team of about 20 people. He says he follows the German automotive market closely because he expects it to go through major change, and warns that Germany needs to move quickly on AI or risk repeating its late reaction to electrification in the old automotive industry.

Quotes

In essence we're a software company, we're not a hardware company.

Jaap Zuiderveld

The most expensive good you have in a company is probably the people who are doing data mining.

Jaap Zuiderveld

Questions answered in this talk

Is NVIDIA a hardware company or a software company?

According to Jaap Zuiderveld at Rise of AI 2022, NVIDIA is in essence a software company: of its 24,000 employees, 22,000 work in engineering, and 70 percent of those do software engineering rather than hardware engineering.

What is NVIDIA Omniverse used for in manufacturing?

At Rise of AI 2022, Jaap Zuiderveld describes BMW using NVIDIA Omniverse to build a digital twin of its factories, simulating production planning, logistics and global team coordination virtually, and expanding the project to three additional factories.

How did NVIDIA technology contribute to COVID vaccine development?

Jaap Zuiderveld says at Rise of AI 2022 that AstraZeneca used an NVIDIA-built supercomputer in Cambridge to develop its COVID vaccine, accelerating the process by more than a year.

Mentioned

Omniverse, digital twin, CUDA, federated learning, Modulus, Clara/MONAI, Riva, Maxine, Merlin, Metropolis

Keep exploring

More from Rise of AI Conference 2022

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 3,781 words

Show the full transcript

now we have our next speaker yap studerfeld and i would already like to invite you to our stage here yeah um we have another yeah we give it another try maybe maybe now it works and um yeah let's go home so first of all thank you um rachman by the way if you're looking for a job join us please i love the way you promote us by the way in a certain way because everything you talk about is what we believe is happening um in the world yeah and uh and for sure um because i i said it today i think you know we need to start working on it not talk about it but working on it and i think in my view um we're talking too much about it but doing it that that's what needs to happen but anyway um i would like to start with a small video about nvidia because a lot of people know us but don't really know us so i i i really want to start with

a small video let's see if this is going to work guys because of the previous presentations that didn't work so i'm gonna click ready yes once upon a time it was all fun nvidia started off by making chips for video game machines graphics became serious business when people started using nvidia for blockbuster movies medical imaging devices and the world's most powerful super computers and then one day researchers discovered that our technology was perfect for ai today nvidia is the engine of ai engineering the most advanced chips and systems and software that makes them sing so robots can lend us a hand cars can drive themselves and even the earth can have a digital twin we live to tackle the world's biggest challenges and don't worry we still love our fun and games ready go sorry upon a time help how do i go to the next one can you click to the next one please can you help me oh sorry the other one so um yes nvidia most of us know nvidia from gaming

yeah now i want to start with that also i'm not here don't worry i'm not going to talk about gaming also but i just want to make sure that we understand it gaming is probably one of the most challenging computational tasks there is in the world a lot of stuff you do today with ai comes from gaming i can tell you it's not only what we do as nvidia but it's also what's happening in the market yeah now i'm not going to talk about speeds or feeds today what i'm going to talk about where are we today where is nvidia today and how much do you really know about us so in essence we're a software company we're not a hardware company we have 24 000 people today 22 000 people work in engineering 22 000 people we are an engineer led company we're not a sales company we are led by engineering out of those 22 000 people 70 do software engineering and 30 do hardware engineering that's nobody really knows us but this

is so important and that's why i'm starting with this slide because that's where we are today this is the full stack of nvidia the full stack but i'm going to start at the top if i look at modulus modalis is a physics-based simulation so what we did let me just explain that what we did we trained models yeah we pre-trained models to make it easier for you to work with our technology and our technology should be available to everybody that's how we look at it we call it democratize our technology yeah we have pre-trained models and sdks to make it easier for you so if i look at molars for example is a pre-trained physics-based model and just as an example siemens energy uses it to develop winter brands wind turbines sorry that's my dutch confused with portuguese and everything else wind turbines clara monae yeah that's a healthcare p-trained model this is a model which we developed together with key synthomas hospital and nhs in the uk which is an open source model to

understand image recognition yeah and we use we use federated learning very important something we can use for example in finance also it's all about protecting your data as you know healthcare data is super important a pre-trained model which you can already use riva yeah reva um riva is a i call it natural language processing model a good example is amex rmx uses it i always like to use examples they're they're um inside their people who take the phone use that model so when somebody calls they can recognize a voice or they can recognize the question and they can answer their customers much better maxine everybody uses maxine here today probably everybody use why i'll tell you why maxine is about enhancing your video images if you use teams if you use zoom you use maxine it's about enhancing your video making it nicer taking the background away you can put a picture you know you know if you use teams or zoom what you can do that is maxine nemo very important nlp natural language

processing but this is really advanced this is where you know as a good example we use alexa where you say something you speak and it answers you it understands what you ask and how you answer a pre-trained model yeah a very important one merlin merlin is a recommendation engine super important i i really don't understand that if you look at the markets in the german market there's not asked a lot about it but these these these are pre-trained models where you can compete against the the guys like amazon for example you don't need to train a whole model you can already take that model and train the bit yourself you can take it and learn and learn how to do a recommendation engine so important avatar you have seen it here we use it here the avatar is creating your own avatar i've seen in a presentation before is where you want to be more human where you create a human which speaks to human i know we are working with different companies around the

world on creating an avatar even within the finance industry for example but in every industry people are interested in it drive is very close to this home market germany it's not home market but to the german market daimler is the best example probably a lot of you know that we have a partnership with daimler we develop the software which they're going to use for the self-driving cars which they're going to put on a road in 2024. they use drive isaac um isaac is all around um sometimes i need to think myself i need to help a little bit it's all about robotics different examples amazon i always like to use an example understanding how you can mitigate the huge centers they have amazon as all of you know have a huge um how do you say that um sorry i forgot the word um uh retail centers but um sorry i'll i'll lose it but it's understanding the automation of their distribution centers sorry i have it again the distribution centers understand what the best

routes is but one of the biggest things amazon had they had a lot of accidents in the distribution centers how to mitigate that and the automation they had in the centers is very low so they were looking with them how to work on that one metropolis is all around video surveillance that's the most important one most of the airports today use metropolis most of the airports today scan already when you come in when you come from a plane they already know before you go out especially if you go to countries like dubai or saudi most of the airports use our technology and you already they already know exactly when you come in who you are what's happening what's doing because they they already have scanned you and last holus holoscan whole scan also very important it's used in medical imaging and it's used for medical instruments siemens is a is a great example siemens is a huge use of holoscan it's all about understanding you know how you develop models which are better pre-check models

which everybody can use it's very important to understand that now i'm putting the full scale here this is you know here we start with what we started with the gpu cpu we're coming with the cpu also the dpu all that is to accelerate your workloads in a data center understanding how a data center works how your applications work one of the things we do is understanding how to approach a market also so we work by industries industry verticals understanding and industry vertical taking people from the industry vertical on board who work for finance who worked for finance previously who worked in healthcare previously and helped us understanding those verticals better and working very closely with those customers and ruckner said it very rightfully how we approach it very fast we develop i mean our route to market is probably five times faster than anybody else because we work together with the customers to develop a product the product is not even finished i know people who join us always say yep the product is not

finished i said so what's the problem we're going to finish it with the customer because then we have the inputs from the customer and we know what's happening it's a i know it's a silicon valley based way of approaching a market but it works i can tell you now to give all this a little bit more i'm not sure if i'm going to do the right one no i went too fast i need to go one back oh no i'm right sorry now put this a little bit in perspective so what's really happening yeah now in fact if you look at what's happening to the acceleration or the speed of where we're able to let's say go for compute task is increasing enormously if you buy today an hpe server with a hopper with hopper architecture which will come out soon yeah with one one gpu that's already faster than a fastest supercomputer worldwide 10 years ago that's one server it's faster than the supercomputer 10 years ago yeah that tells today we have three

million developers worldwide working with us now the three million is one but do you know that more than half of them joined us in the last year yeah more than half of them joined the last year that is important to understand now this will you know we have we are following 10 000 startups worldwide by the way 450 we need to talk we need to synchronize a little bit 450 in germany 415 germany those are people who only work with our technology because sorry that's what we do but it tells a little bit about what's happening and i i can i can echo ragna the majority of our starters are in israel now israel is also one of our home markets as we bought mellanox two years ago when the deal was endly closed we have about three thousand people in israel and the start of the community there is thriving seriously you learn so much from the startup community and we work so closely with that together today we accelerate more than 2700 applications

this is really important these are applications which are you know applications which are put on the market by by isvs independent software vendors now you can say to me what does that really mean speed up or accelerate what why do i really care because it's working now i will tell you why why it's so important there are many examples but if you want to have an outcome so a lot of tasks today because the data is becoming a quiet um let's say bulky and if you want to work on on on a model sometimes you need to wait a day before you know the outcome of a model or before you know an outcome if you have the answer within five minutes it changes your business perspective completely you make decisions in a complete different way you're much more agile you're quicker you're faster it's not replacing but it's you're able to make a decision quicker that's why acceleration is so important and you're able to do much more work also you know what this

is always interesting the most expensive good you have is your people the most expensive good you have in a company is probably the people who are doing data mining they're the most expensive ones because they are rare yeah and you give them a system which they need to wait so they go for a coffee they wait you know what the cost is of a data scientist a really good one at least half a million at least a total cost then you're very cheap but it goes up if you have a lot of them and they're waiting before they can do anything it doesn't matter the cost of hardware is just doesn't matter so it's really important to understand that today we we we have more than 100 450 sdks and models we support and we have more than 8 million cuda developers this is really important this is our bella programming language which is used on a gpu and i think you know going to the next slide how do i capture this a little

bit um you know for me ai occurs everywhere that's what i said in the beginning yeah for us as nvidia we have a growth of 50 to 60 percent year over year yes also in europe i have people asking me also in europe because it's easier because the number is lower so it's easier but i can tell you it's also happening in china it's also happening it's happening globally it doesn't matter yeah but a huge opportunity yeah here went through the numbers we have today 25 000 customers 20 globally working without technology we have more than 1300 partners understanding our technology and here i mentioned some of the industry leaders which rely on us european the couple i always like ek as a very good example they are a big fan of omnivorous yeah daimler everybody knows bmw omniverse i'm gonna show that a little bit siemens uses in many aspects red bull you know the reason that my fellow country man max verstappen is able to win is because we help them designing a

better car we're a little bit involved it's interesting but you know these are things and i can go on and on and on and on if i only look at germany there's so many initiatives happening i mean i know that because i'm talking to applied ai i'm talking to dfki what these guys are doing is just but they need more money they need more hardware to do the work and and there's so many opportunities it's it's just immense i i can tell you um i mean this one is interesting also astrazeneca you know that astrazeneca we built a supercomputer in cambridge and they used that supercomputer to develop the vaccine for covid they accelerated by more than a year to enable them to to produce a vaccine which which came much faster the market all things which are really relevant i mean for me um what i always like is the human brain project which we're doing in ulif to understand how to cure parkinson's that they use our technology also in eureka but very

important how disease starts because then you can cure a disease things which are all really important now for me the next wave will be um i call it um understanding um the 3d world as is today in a virtual environment yeah what will happen is if we look at architecture if i look at design everything else is how do you simulate a physical base real model in a virtual environment so important that will save so much money that's going to be the next thing that's going to drive the next thing it's going to be very important for media entertainment today more than 8.6 billion of compute hours is used for rendering in media entertainment it's immense these are all things we we don't really know about but it's so immense and doing realistic base rendering is something we're able to do product design manufacturing robotics autonomous vehicles all these guys work in in virtualized environments to understand their products better to save money while they develop their products and i i can tell you there

are many examples who are already using this technology this will certainly be the next wave now i'm going to use one example i need to look a little bit at times sorry this is bmw bmw is a is a very well known example i think we talked about it but they built a complete digital twin and where they simulated um how to build a factory or how to set up a factory effectively yeah how to plan it how to get all the people worldwide together in a in in in a virtual environment it's it's just um what happened there is quite unique i know i i will show the video a little bit it doesn't really speak but when you really understand what it means there's a lot behind it there's a lot when it comes down to logistics how to improve logistics how to improve costs where it went wrong they could all understand it without the people being there in a digital environment i'm gonna use this video bmw has continued building on

the momentum of their factory of the future initiative we showed you earlier this year on average they produce one new vehicle every minute expanding the initiative to now include digital twin projects at three additional factories which total more than six million square meters meeting bmw's demands for continuous improvement and innovation requires simulation of complex production scenarios to speed output increase agility and optimize efficiency everywhere nvidia omniverse introduces new simulation capabilities everywhere bmw needs to make critical decisions and introduce automation bmw and their partners have also been connecting new users and tools to expand the omniverse ecosystem spanning every aspect of the bmw factory of the future omniverse allows bmw to collaborate and simulate from the entire plant level down to comprehensive engineering detail it's quite special i can tell you sorry i'm gonna make the mistake again can you click to the next slide so for me i just want to close down on a couple of words we we close very close partner with also with hpe yeah now how i approach always this

is we're there to help hpn to help you any questions let us know talk to us if you don't understand it because it's complex what we do ai is complex it's not easy i can tell you i'm now working eight years in this business i was there right from the beginning when we started it is complex but there is a lot of things you can do the breadth of technology we have on board is enormous i'm more than willing to share with you for you to understand how to work with us together with hp more than welcome i'm going to close with a small video around ai i always like that sorry it's a or i can just stop i can i can just stop and leave it that's maybe better i'm not going to close because i'm all over time already i just had a small video around uh you know ai but let's leave it it's fine thank you very much for your time i could spend much more time here but thank

you very much yeah thank you i think we could watch uh also some others you have so much uh with you uh really really cool and so i know you are a dutch living in london but you are you also want to take care of germany um as ragnar also like to do that um and i would like to um know how do you do that how do you represent nvidia in germany yeah so we we have a so i i'm always from principal uh working with local teams i have a local team of uh well we have in fact almost 1 000 people in germany yeah but that's yeah that's a lot of software engineering in munich and in berlin by the way because in berlin is our headquarters by the way on development on omniverse yeah okay so we have there's a lot of knowledge in this country available but i have a commercial team of about 20 people who um all live here and and talk to you there are a lot

of people here in the room now from my team um you can put up your hands uh if you see just come to us no i work from the principal with a local team and we support the local team but i'm really interested in in what's happening in germany for many reasons i believe the automotive market still is is a is a huge market which will go through a huge change so that's very interesting how to follow that and for the rest um you know i i think germany has uh has a huge potential um but but they need to get on it if not the same thing will happen what happened in the old industry with electrification they a little bit kind of were too late and here they need to jump on it that's my my my personal view so yeah yeah i would like to discuss some more questions but um yeah we are already out of time so thank you so much for your presentation and we'll see you later thank

you

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.