Jon Krohn: 00:00
This is episode number 543 with Nicole Büttner, Founder and CEO of Merantix Labs.
Jon Krohn: 00:10
Welcome to the SuperDataScience podcast, the most listened to podcast in the data science industry. Each week we bring you inspiring people and ideas to help you build a successful career in data science. I’m your host, Jon Krohn. Thanks for joining me today. And now let’s make the complex, simple.
Jon Krohn: 00:41
Welcome back to the SuperDataScience podcast. I am oh, so very excited to have the eloquent and effervescent Nicole Büttner for you as our guest on the show today. Nicole is the Founder and CEO of Merantix Labs, a renowned Berlin-based consultancy that enables companies to unlock the value of AI across all industries. She is a member of the management board of Merantix Labs’ parent company, which is conveniently named Merantix, and that’s an AI venture studio that has raised $30 million in funding from the likes of SoftBank to Serially, Originate’s successful machine learning startups. She holds a Masters in Quantitative Economics and Finance from St. Gallen University in Switzerland, which is the world’s leading German language business school. And she was a Visiting Researcher in Economics at Stanford.
Jon Krohn: 01:32
In this episode, Nicole details what an AI venture studio is and how she founded a thriving AI consultancy within it. She talks about how to spark AI innovation in a company of any size, how to make great use of the unlabeled, unbalanced data sets that abound in business, how to engineer reusable data and software components to tackle related projects efficiently, and the three distinct types of founders she looks for when she puts together the founding team of an AI startup. Today’s episode touches on a few technical details here and there, but the episode will largely be of interest to anyone who’s keen to make the most of AI innovation in a commercial organization, whether you happen to have a deep technical background today or not. All right, you ready for another inspiring episode? Let’s go.
Jon Krohn: 02:25
Nicole, welcome to the SuperDataScience podcast. I have been excited about this episode for a long time. You have such an amazing perspective on commercial applications of AI. And I can’t wait for the audience to hear everything you have to say. First off, where in the world are you, Nicole?
Nicole Büttner: 02:46
Right now in this very moment, I’m actually sitting in Switzerland, but actually the company is based in Berlin. So I spend most of my time working out of Berlin.
Jon Krohn: 02:55
Nice. And it’s too bad… I think even in the video version of the podcast, the view over your shoulder will likely be cut out, but it looks magnificent. So I could see why you are spending time in Switzerland while you can. So I know you through Rasmus, who is the Founder and CTO of Merantix, which we’re going to talk about a fair bit in a moment. And I’ve known Rasmus for a long time. So when he was doing his undergrad at Oxford, I was doing my PhD and we were both members of the Oxford Entrepreneurs’ Student Society. And so I mean, that’s more than a decade that we’ve known each other. And every once in a while, he’ll be in New York and we catch up. And every time we catch up, I’m like, oh my goodness, I can’t believe the another enormous order of magnitude increase in success that he’s had. So if people want to hear from Rasmus directly, prior to hosting the SuperDataScience podcast, I piloted my own podcast called the Artificial Neural Network News Network, A4N, and Rasmus was a guest on that show. He was on episode four. So you can hear from him.
Jon Krohn: 04:15
But this episode is all about Nicole. So Nicole is on the management board of Merantix, which is a super cool company. It’s an AI venture studio. And Nicole will be able to provide you with more information on exactly what that means. But the idea is that you incubate companies from scratch. I think even a lot of the time you help create of founding teams and you help them figure out what idea they will be working on. And to give a little bit of context on that, examples of very successful ventures that have, in their own rights, either raised tens of millions of euros in funding, or even been acquired, include Vera Healthcare, which does machine vision models that can detect tumors in, I think originally it was breast cancer, but maybe it’s other cancers now as well. You can speak to that when I finally give you a chance to speak. And SiaSearch also works in machine vision, but they were labeling data that comes from self-driving cars. So there’s tons of sensors on self-driving cars that gather tons and tons and tons of information and so being able to label those data and find pedestrians or stop signs and that kind of thing.
Jon Krohn: 05:33
So these two separate ventures SiaSearch, which has now been acquired by Scale AI, Vera Healthcare, which is doing extremely well as a standalone startup, this is a perfect example of how Merantix can be hugely valuable to incubating these AI ventures because they both involve machine vision. So this gives a perfect example of how two different kinds of companies with different application areas have overlapping underlying R&D; they’re both involved in machine vision. So in terms of how we engineer this, how we scale it, there’s common knowledge that can be shared. And this shows one example of how this AI studio model can be so helpful. And clearly it’s doing very well. Those kinds of early examples like Vera Healthcare and SiaSearch having such early success, including that acquisition, has led to lots of funding. Now $30 million in funding, including investment from SoftBank, one of the biggest names in venture capital investing. So that gives you some context on what Merantix does. And alongside all of those ventures, there’s also something called Merantix Labs. And Nicole is the CEO of Merantix Labs. So Nicole, I’ve been speaking for way too long, trying to give some context. So one, feel free to correct any of the mistakes that I made or add a bit more color on the Merantix ventures, and then tell us about Merantix Labs and what you do.
Nicole Büttner: 07:11
Sure. Thank you very much, Jon, for the invitation. Really happy to be here. And I’ll talk a little about Merantix. And I agree; actually, you’ve raved about Rasmus and I think that was also one of the reasons why I joined Merantix, to be very frank, because the quality of people is just extraordinary. Also Adrian, who’s Rasmus’ co-founder, I want to say at that stage of my career was really looking for the quality of people I would be working with curious minds, who are brilliant in their respective fields. And that’s what I really found at Merantix. And happy I took that decision two and a half years ago.
Nicole Büttner: 07:52
I think you’ve already given a pretty good overview of Merantix, but I mean, we’re basically looking to really create impactful applications from AI. There’s a lot of AI research going on and we also do some of that at Merantix Labs. But for us, the main driver is really to create commercially successful companies that have real world impact. So to make this technology useful, I would say, for mankind, for industry, for everybody, in the economy and society, and that’s really what our ventures do. And you’ve mentioned some of our ventures that are more advanced. And we’ve also incubated some new companies in very interesting fields. One, for example, is [Brink 00:08:41] in the ESG compliance space, more an NLP focused company, for example, a focusing more on language. And another one would, for example, be Terra [Lumina 00:08:51] in the nutraceutical space. So really let’s say more a drug or supplement discovery space to understand what are the powers that line the plant universe and how can we leverage the best for mankind.
Nicole Büttner: 09:08
So there are really a lot of topics that we think about. And I think the common denominator’s everybody’s very curious, people are excellent in what they do, and they really want to create impact with their companies. And that’s how I then entered Merantix as a founder as well. So I joined Merantix two and a half years ago as a founder and created Merantix Labs. So that was really sort of my story. And I would say Merantix Labs is a little bit different probably also from Merantix. [inaudible 00:09:46] ventures because we’re, first of all, more of a subsidiary. So we’re not raising any more capital in the market. We’re profitable. And also we’re an integral part of the studio and also we’re a service provider. We’re not building a product, but we’re basically taking this knowledge of data driven, AI driven, business model that we have from creating ventures and infusing it into existing organizations and companies that exist. So we’re working with big fortune 500 [inaudible 00:10:19] companies, as well as world champions SMEs. We have a lot of those in Germany and that region that build components that go into products that you know from everyday life; companies you’ve never heard of.
Jon Krohn: 10:34
Yeah. So I didn’t know that piece, that Merantix Labs was kind of… so the Merantix AI venture studio has created lots of these ventures. We’ve named some of them; Vera Healthcare, SiaSearch, Brink, [Cambrium 00:10:47]. And then Merantix Labs, that was initially just another kind of venture that you were the founder of, but then now it has this special status within the firm. Whereas the other ventures attract outside investment or could be acquired, Merantix Labs is this wholly own subsidiary focusing on AI R&D and operationalizing AI for corporate clients. And amazing to hear that you’re already profitable. That’s super cool in such a short time.
Nicole Büttner: 11:19
Hmm. Yeah. And I would say the reason why this fits into the universe so well is that we are trying to get together really good research and make it industry- relevant. So Merantix Labs sort of builds this bridge into industry and is also the vehicle through which we conduct corporate R&D, but also some research projects with excellent academic institutions. And that gives us insights that inform, okay, these are the type of problems that industries have, and that industry in general has, and maybe we need to develop some solutions or ventures around this. And the other side, we obviously bring in the founders and some experts in some fields. And that also is very beneficial for us to kind of understand certain spaces better. So it’s a really nice synergy and that’s why we made it more permanent, I would say, fixture in the Merantix universe.
Jon Krohn: 12:10
Totally. I can see that they complement other very well and they complement each other so well, in fact, and you’re doing such a great job running the Merantix Labs’ subsidiary that you’ve been invited to be on the management board alongside Adrian and Rasmus of the broader Merantix AI venture studio. So congratulations, Nicole.
Nicole Büttner: 12:34
Thanks.
Jon Krohn: 12:37
The thing that I saw on LinkedIn that immediately compelled me to reach out to Rasmus and ask if he could introduce me to you and we could have you on the show was because I saw that you’d given this talk on how to spark AI innovation, which clearly was you are a global expert on. So help us with that. So whether you are a company that maybe just has some data that you’ve collected, and you’re not sure how to innovate with machine learning with those data, or maybe you’re even earlier stage, and you’re not even systematically collecting your data in any way, how can companies be sparking AI innovation systematically?
Nicole Büttner: 13:21
Yeah, I mean, I think that’s a question that’s sort of a universal question almost in organizations of different sizes. And I think also kind of a key question, if you want to stay relevant as a business. Because it also asks the question, how can we use AI? For me, it means how can we use AI to transform our business model, create new products and services serve our clients better, et cetera? So I would say there are like two, three points I want to stress with this. First is, start with the impact you want to have. So I know we’re like the AI shop and we will look at it from that AI angle. And we will tell you, is this an AI a problem or not? Can we solve this with AI or not? But I think the main thing is really what kind of problem are you looking to solve? Which kind of client are you looking to serve? What is a solution you can envision? Then AI can be an instrument.
Nicole Büttner: 14:03
I think it’s no use to try to implement AI just for the sake of it. I, myself, I’m an economist by training and not an engineer, computer scientist. I believe in technology as an instrument, and that’s also the way I look at AI. I think that’s also how organizations should look at it. I think a second really essential point is, don’t be too afraid of trying things out often. This might be a very German perspective. I don’t know. It’s an engineering-heavy country. But often I hear, “We need to build first all the architecture,” the whole data architecture of the house, “and then we can think about use cases.” I always think, “Yeah, no, you’ll never be done building that house,” because even tech-first companies, they just tear it down every few years and rebuild it because technology is advancing so fast. Maybe let go of this notion that you can create structures that will last forever. It’s a fast-paced environment. I think personally creating, even, data infrastructure, data architectures without knowing the use case is pretty obsolete, because it will determine what kind of architecture and infrastructure you actually need to serve this use case. I would say, be open to this being an ongoing process. That’s maybe something, if you’re used to building a project and designing one model, and then go, “Okay. Now we mass produce,” that might be a new notion for you, but …
Jon Krohn: 15:47
Right.
Nicole Büttner: 15:48
That’s … I have to think about. Then the willingness to experiment with that, I think, is really, really key.
Jon Krohn: 15:59
Many thanks to the Master of Data Science program at the University of California, Irvine for sponsoring today’s episode. The UC Irvine Master of Data Science program blends statistics and computer science principles with partners from industry to empower students to innovate in the field of data science. Located on the tech coast of Southern California, students of the program will enjoy a powerhouse ecosystem where over a third of Fortune 500 companies are located. Take a giant leap in your data science career through the UC Irvine Master of Data Science program. To learn more, head to SuperDataScience.com/uci. That’s SuperDataScience.com/uci. Check it out.
Nicole Büttner: 16:46
I think, thirdly, it’s really a matter of culture and safe space. We’re in the deep tech sector, and I like to say the more deep tech I go the more touchy-feely it gets. Because, this technology sparks many fantasies, and also a lot of dystopian fantasies of people. The fear of being obsolete, or, “What is my purpose here?” And so on. Organizations are ultimately … We’re all people. We all, also, in little things, have resistance to change. That’s just how we work. Even I have to change my brand of, I don’t know what, coffee, I’m like, “I don’t like that.” I think being aware of that, and really understanding that you have to create a safe space for this innovation as a leadership team and as a whole organization, and that it’s also a cultural transformation for most companies. Especially here in Germany, a lot of companies have been around for centuries, and have developed very high expertise and excellence over centuries. That’s a cultural shift, and you have to embrace it and create a safe space for it.
Jon Krohn: 17:59
Cool. To spark AI innovation, three points from you, Nicole. We want to identify what problem we’re solving. AI could be an instrument for solving it, but maybe there is some other solution out there, or some complimentary solution. The second thing is to not be afraid to start or experiment, because we can’t architect the perfect solution before we’ve started anyway. Then the third one is to create a safe space for innovation. I love these. I love in particular the second one. If you want to spark AI innovation, you have to start doing some innovation. Perfect. In terms of a specific example of a recurring problem that you help clients out with, I know that you often come across small, unbalanced data sets that need to be labeled. A big part of the R&D function at Merantix Labs is using unsupervised learning models to solve these data labeling issues. Do you have any, maybe, interesting case studies, obviously without divulging proprietary information, but some kinds of examples that add color to that use case?
Nicole Büttner: 19:17
Sure. I would say the problem is one that … It goes back to the way industry’s structured, maybe, in Germany, in particular in Europe, but also something that also bigger corporates come across. If you want to build machine learning applications, often we’re talking about supervised models. You label data sets. It all looks great in academia when you study the field. You think, “Here I have my class, this class, and then I train a model.” Then you look at reality, and you’re like, “Oh, my God. This looks horrible.” It’s way smaller than people think. People are always saying, “Oh, no, we have enough data.” It’s always less than they think, it’s always less balanced than they think, and it’s always a pain labeling it in the end, ultimately.
Nicole Büttner: 20:01
We believe, trying to create really scalable impact, we have to provide … This is a really universal problem, large or small organizations. We have to provide a solution that addresses this. Our solution conceptually is to say, “Okay. We want to build big models,” because also the bigger, the more parameters and data go into a model, the better it is, just in a very simplistic way of looking at it. We want to build big models that perform really well. We want to use unsupervised models, because we just have much bigger unlabeled data sets than labeled data sets.
Jon Krohn: 20:42
Totally.
Nicole Büttner: 20:42
Usually. Then we want to use these unsupervised models and fine tune them with labeled data sets to very domain-specific areas. I think that’s a very powerful idea, and that’s methodologically not completely solved yet in all areas. But, we have our brilliant VP of machine learning research, Johannes, who came from Open AI and other big shops to us to lead this effort, and who is doing a really great job to develop a lot of the methodologies we need to create this. This is, for example, something we do in the language space. Talking about language models, you obviously have a huge corpus of text available in the internet. But, when you think about more domain-specific applications, and that’s usually what corporates need, they very rarely have use cases where you just use the internet. It’s usually in the legal field, or you have something that’s very specific to creating certain offers for clients, or to finding details in some of your agreements, et cetera, or workflows. Then you basically need something like this to understand this.
Nicole Büttner: 21:58
I think this is one of the powerful applications in the field of language. Then we are also looking at this, for example, to serve SMEs because they’re … In addition, you have a lot of companies that are very siloed and small, so per se have smaller data sets. We’re really looking for scalable models to serve these companies and to bring them into the 21st century. For example, one of the use cases we’re working on is in computer-aided manufacturing. One of the processes here in Germany, we have a lot of machinery producers and so on, is, they basically get a sketch, which is a computer-aided design, from one of their clients to say, “We need this component when building this car.” Then they have to try out on a machine how to make this part, this prototype part, from, I don’t know, a block of metal. Have to remove some parts, for example. It’s a subtractive method.
Jon Krohn: 22:56
Yeah. And sculpt it.
Nicole Büttner: 22:57
Yes, exactly. This can take many-
Jon Krohn: 23:01
Like a block of marble.
Nicole Büttner: 23:01
Exactly. This can take many iterations, and obviously is time consuming. You need a specialized engineer doing it. It’s also resource consuming, ultimately. To make these iterations faster, it’s really useful to understand which are general rules to translate these computer designs into the manufacturing instructions, because you want to basically program a machine to then build this part out of this metal block. That’s a subject that many manufacturers deal with in Europe and in Germany specifically. Some of them come compete with each other, but not necessarily all of them. You can be building many parts. Where this same logic can be very powerful, because you can train bigger models on this fundamental mechanic, and then fine tune it on a more specialized area with the labeled data set, just become that much more impactful and faster to serve a client.
Jon Krohn: 23:58
Perfect. Yeah. I understand completely. This is definitely a super powerful approach that you’re taking, where you have these very large unsupervised data sets, like you give the example of … You could use all of the internet, so you could use all of the English on the internet or all of the German on the internet, or maybe even a combination of languages. All of the language on the internet. You can pre-train a gigantic language model to understand, in a way, to have a representation of the meaning of words, and maybe even be able to translate between the English version of a word and the German version of word. Then we take that gigantic model that was trained on a very, very large unsupervised data set that doesn’t have any labels, and then you can fine tune it to some specific domain-specific problem that one of your clients is looking to solve. They might have a relatively small amount of label data that you can use to fine tune, and then validate that the way that you fine tuned it to their problem will be effective in practice.
Jon Krohn: 25:03
Amazing, Nicole. That was a really cool example of a specific R&D problem that you’re solving. It sounds like, with that kind of problem, there would probably be some components that you could reuse between different projects. You could have a giant language model, say, that hasn’t been pre-trained to any particular purpose, and then that could maybe be reused for multiple different projects, fine tuned to multiple different, specific-domain, smaller data sets. Do these kinds of reusable components make it easier for you to tackle projects and then deliver your projects efficiently?
Nicole Büttner: 25:51
For sure. I think that’s a really important topic you’re mentioning here. We’re essentially a project organization. We build cool code for other companies. This has a project character. But still, we have an AI DNA, so we don’t want to be manually doing the same stuff all over that we can automate or build and repurpose. I think a good example for, really, excellence and delivery, and also to build these scalable models where you can deliver impact fast, is when we go back to computer vision. I think if you have some of these open-source libraries and computer vision, pretty much you can probably build a classifier from anywhere in an internet cafe using some of these libraries. But then, if you have more specific problems, it’s really big images, and you’re looking for really small things in those images, then maybe some of these libraries are not so effective.
Nicole Büttner: 26:44
But, still, these problems are quite common, again. Three of these domains, for example, that we would tackle with similar components … We like to call this particular component “chameleon,” for example, in our zoo of models. Is we can use, basically, in cases from medical imaging, when you’re looking at really high resolution pictures, and looking for really small cell defects or anomalies, as you do in cancer detection. But also, if you look at, for example, satellite images, when you’re looking at all the earth observation applications where you have really big images, and you’re also looking for pretty tiny stuff. You’re looking for a specific crop defect. You’re looking for a specific demarcation. You’re looking for, maybe, a person moving, or a boat, or something like that. That’s a very similar problem.
Nicole Büttner: 27:45
Then, jumping to a third application, for example, is damage recognition, when … For example, we also built a model for a company here in Germany, and it’s all about returning cars, basically, and damage estimation on cars. And there you have the weirdest perspectives and everything, and ultimately partly tiny scratches or dents from hail or whatever. And you still need to work with this information in small pixels. Also something in autonomous driving.
Jon Krohn: 28:17
Right. Right.
Nicole Büttner: 28:17
When you look at very small pellets on the road hat take up very few pixels of image. And so this, I think, is a really good example for building reusable components that go beyond your open source libraries that are quite commoditized, but that are still very useful and transferable across domains and that can be fine tuned on specific domains. And then obviously how to fine tune that and so on, that’s a lot of the secret sauce that our engineering department has figured out. But I think those are quite powerful applications.
Jon Krohn: 28:53
Cool. Thank you, Nicole, for so many rich examples that illustrate this idea of reasonable components and how they make delivery of projects more efficient.
Jon Krohn: 29:05
Struggling with broken pipelines, stale dashboards, missing data? You’re not alone. Look no further than Monte Carlo, the leading end to end data observability platform. In the same way that New Relic and Datadog ensure reliable software and keep application downtime at bay, Monte Carlo solves the costly problem of data downtime. As detailed in episode number 499 with the firm’s brilliant CEO Barr Moses, Monte Carlo monitors and alerts for data issues across your data warehouses, lakes, ETL, and business intelligence, reducing data incidents by 90% or more. Start trusting your data with Monte Carlo today. Visit www.montecarlodata.com to learn more.
Jon Krohn: 29:55
So going from those specific examples to something more general, I know that you have something called an AI canvas approach. It’s a model for successful AI transformation. Can you tell us about this AI canvas model?
Nicole Büttner: 30:11
Sure, Jon, So this is back to sparking innovation and starting an impact. I think we developed the framework together with the University of St. Gallen with basically in mind that… I’ve been in the sector for like, oh God, I hate to say this, almost 10 years, but the evolution has been some companies are obviously AI natives and they have had this in their business model from the very beginning. But most organizations, quite timid, a lot of them are still not really using the technology but anyway, starting with prototypes and first it tries to experiment with the technology. And what we realized is that a lot of these organizations have difficulty then scaling this application into really powerful organization-wide, real life application that also can unfold impact and give some return on investment and not just basically some investment.
Nicole Büttner: 31:08
And that was the very first starting point of starting this canvas and say, “Okay, what are actually some of the key components and areas you have to focus on, even at the very beginning stages just so you know what’s coming along the journey if you want to build a scalable AI application?” And obviously, when you’re building the first prototype, you don’t have to think about all the legal ramifications, but it’s good to know about them because maybe that’s ultimately a show stopper and you don’t want to go down that road. So we’ve basically created a framework that’s divided into different sections. One is a business section. So very much focused on the pain point, the impact you want to have with the solution you’re building.
Nicole Büttner: 31:48
And it sounds very trivial. But really it’s worth thinking about this a lot. Because I’ve seen many applications which sounded great and we read about AI and then we created this use case and ultimately management like, “What kind of problem are we solving here? This is not relevant for our organization.” And then you lose buy in and people don’t get promotions for it and leadership is not interested. And then it dies down slowly. And I think nobody’s motivated to build these kind of use cases. So impact is key. I cannot stress this enough. That’s the first. What kind of invest do you want? What could be a way to actually monetize this then again? How does this change your business model? All these questions go into this first part.
Nicole Büttner: 32:28
The second one is focus on organization. And organization goes to the fact that when you create machine learning products and services, you need an ongoing process. Because as you know, models don’t magically learn themselves and become better but you need to monitor them. You need to make sure that they’re still doing what you want. This needs to still be aligned with your business purpose, et cetera. So you also need quite efficient and effective ways of organizing people that need to take decisions on which use case to prioritize, but also who’s taking responsibility for these models and how do we take decisions? And I think that’s also often the component you think about as a business to create these, I don’t want to say agile, but dynamic structures that correspond to also the technology.
Nicole Büttner: 33:18
The third component is more focused on data and technology, so what kind of data do you need, data architecture. It’s focused on your tech stack, which tools do you use. It’s focused on things like compliance, safety, robustness, other regulatory concerns. And they’re also very important to keep in mind and to understand what are also some of the risks, some of the fields we need to master. And the fourth, tying everything together is this AI life cycle. Just to stress, it’s an iterative process. It’s not, “I build something, I install it and then I look the other way. It will probably run and maybe I need to do few buck fixes.” It’s a constant, incessant process, so to speak that you need to stay involved with. And I think that’s something that makes it quite complex, especially because the technological advances are so fast and you can make sure that your whole calibration is still up to date and that you’re still using the right components, et cetera, et cetera. Regulation, everything is changing quite quickly.
Nicole Büttner: 34:28
And yeah, I’m very happy to say that we’ve basically managed to train a lot of companies on this framework. A lot of companies just go to our website and download it and use it as a first orientation guideline. But we also implement whole AI hubs within organizations using this approach. And I’m really happy that we could develop this with the University of St. Gallen because you, they have academic excellence. We chose a business school for a reason, right? We also have close ties obviously to engineering schools. But this was really for us quite key to work with a school that has excellence in management and business impact because that’s ultimately what we want to create. So we wanted to give a toolkit for everybody from the management and business side trying to approach AI and use cases.
Jon Krohn: 35:24
Cool. Thank you for that explanation of the AI canvas that allows companies of any size to be able to successfully transform their organization with AI. And you talked about the key pieces there, knowing what the impact is going to be ahead of time, figuring out what data are needed and then monitoring throughout the AI life cycle. My next question was going to be if there’s some place that people can get access to this, and it sounds like we can get it right from the Merantix website.
Nicole Büttner: 35:52
Yes.
Jon Krohn: 35:53
And so we will be sure to include a link in the show notes to that because I’m sure a lot of people are going to want to be able to follow along in all of the detail provided. Lots of organizations out there that want to be doing AI transformation. So you mentioned a couple of times there that this was in collaboration with St. Gallen University. And I love that you mentioned them that you’ve been working with them. I have such a strong affinity for this university. So I attended in 2013 and 2014 something called the St. Gallen Symposium, which I understand you have attended as well.
Jon Krohn: 36:28
So you attended it as a student at St. Gallen University. You did your bachelors and your masters degree there in quantitative economics and finance, which is such a cool background for what you’re doing today. And then you were invited back last year to be a speaker at the symposium. So this symposium for me as an attendee was a life changing experience. I met some of the top thinkers from all over the world. So there’s a blend of young people and more experienced leaders, so what they call leaders of today and leaders of tomorrow. And those leaders of today, it could be the CEO of Microsoft or Christine Lagarde. And so these really are, these are the top business and political leaders on the planet go to the symposium.
Nicole Büttner: 37:24
Yes.
Jon Krohn: 37:26
But the unique thing is that young people get invited as well. So these leaders of tomorrow, which I was almost 10 years ago now. And so you get these amazing people, these leaders of tomorrow, also from all over the world. And for me at that time, it was a really eye opening thing. I realized that I could be and should be doing something much more impactful with my career than I was at that time. And so it completely recalibrated the way I was thinking about what I could do with my life. And also the connections that I made there, lifelong friends that I’ve made. So the St. Gallen Symposium absolutely incredible opportunity. And I would love to hear more about you experience there, especially as a speaker recently. But to let listeners know if this sounds incredible to you, and you would like to be invited as a leader of tomorrow to the St. Gallen Symposium, well, you can actually go for free.
Jon Krohn: 38:36
So this might sound too good to be true. So Nicole and I can vouch that this is not some kind of scam. So you can simply submit an essay, and essays are due soon, they’re due on February 1st. But if you are under 30, I believe the cutoff is, and you are a student in a masters or PhD program, then all you have to do is write an essay. And if your essay is one of the hundreds selected, then you get free flights, free accommodation, free attendance, everything free, food, everything at the symposium. And that the idea being that bridges the leaders of tomorrow with the leaders of today. And in case you’re wondering, this is bankrolled by the leaders of today who sponsor it and pay to be there to have this opportunity to meet so many different young people. So again, I’ve now been doing this long monologue, Nicole, corrected me on the things that I got wrong. And yeah, just let us know about your experience at the world’s leading German language business school, as well as the symposium in particular.
Nicole Büttner: 39:43
Yeah, I think what’s quite fascinating, and I was a student at the University of St. Gallen when I first attended, is that it’s all also student organized.
Jon Krohn: 39:53
Oh, yeah.
Nicole Büttner: 39:53
This is basically a student initiative and it’s student led. And from the drivers you meet that pick you up from the airport, it’s super professional. They’re all also students at this elite university and everything is on a very professional and inspiring level. And I think for me as well, what I really liked about the interaction at the symposium last year when I went again and also the first time I went is taking yourself out of the day-to-day business context and thinking about the wider context of what you’re doing. And I think that’s also part of what we try to live at Merantix and Merantix Labs.
Nicole Büttner: 40:33
Eventually you have… AI sparks so many, as I already said, dystopian ideas. And we really want to use AI as a tool to make this world a better place, to make patient experience better, to make customer experience better, to make your experience getting some government service better, to make the manufacturing process more efficient, to make, I don’t know, better, cooler materials that are more sustainable. Ultimately, that’s what drives us. And that’s what we want to contribute to. And I think a setting like the symposium reminds you to stand back and think about, “Ah, am I still focusing on this? How can I with my toolkit contribute to that?” And that’s also what I ultimately very much like about Merantix and the whole ecosystem we’re operating. And we, for example, just opened the AI Campus in April last year in Berlin.
Jon Krohn: 41:26
Oh yeah. We haven’t even mentioned that on there yet.
Nicole Büttner: 41:29
There you go. And that’s really fundamentally rooted in the belief that we need to build ecosystems to create impact and to create it fast. And basically, the AI Campus is this idea… And it’s a not for profit initiative from Merantix, so we organize it, it was a lot of work, but we don’t really make money from it, but we bring together different players that we think need to be in the same room or under the same roof to really create impactful innovation, bring in also our Merantix ventures, for example.
Nicole Büttner: 42:03
But also other ventures, even competing ventures in the space of machine learning, deep learning. We bring in investors, we bring in corporates, and their data or AI teams or parts of them. We bring in the new, for example, spinoff from universities that are machine learning and AI focused out there. We bring in some researchers at time. We bring in government, the first German GovTech Hub has its home on the AI campus in Berlin that Merantix created. And I think that gives you a very good perspective of how we try to create really an ecosystem for innovation because it’s such a growing field and we really need to make sure we get it right and we use it for the right applications. We feel it’s our responsibility to create such a visibility for the field to bring people together and also ultimately to focus it on applications that we think are good applications.
Nicole Büttner: 43:10
And also, quite frankly, the power of a physical space in these times of COVID is probably… Sometimes like people are probably longing, have this weird feeling of longing again for physical meetings, but it’s a physical space in a neighborhood of Berlin. These are real people, these are the people building these applications. So it’s also a place where we take responsibility. We show face. We’re like, we’re these people behind these companies. We’re these people building it. And it’s really in the middle of Berlin and quite open; we do a lot of events. And so I would invite you also to check out AI Campus Berlin, to look at… Some of the events are open for everybody. You can just tune in digitally, virtually, and connect to the community. And I think that’s quite exceptional. And that’s also a huge part of what motivates me to build this company also in this ecosystem.
Jon Krohn: 44:04
Super cool. Thank you for bringing that up. And I can’t believe that it took us that long in the episode to get into the Berlin AI Campus, which is an absolutely brilliant idea. I’m sure you’ll see imitators all over the world. I certainly fantasize about imitating something similar in New York. And I understand from Rasmus that one of the big keys to have having the AI Campus work so well is that you only have one coffee machine across the entire giant building.
Nicole Büttner: 44:32
Huge point of discussion. Yes. But that’s true. Yeah. Yeah. So that’s kind of the watering hole, you would say, everybody meets in the queue, or around the coffee machine. Also tea drinkers are included, of course. And sure, by the way, Jon, if you want to create another AI campus, I think we would be the first one to do it together with you and to sort of even encourage imitation because we really believe in this model, and it’s not a Merantix-only job to create these hubs in a worldwide ecosystem, ultimately. So yeah, I think that would be great.
Jon Krohn: 45:08
Wow. Maybe you’ll start having the gears turning out of fantasy and into reality. That would be extraordinary. Well-
Nicole Büttner: 45:14
Yeah. Let’s do it.
Jon Krohn: 45:15
Whew. Wow. That’s an exciting idea. You heard it here first. All right. So beyond just spinning out entire AI campuses with their lonely single coffee machines, more generally speaking about the Merantix idea of having founders come in and then you help them create teams and find some commercial application and then help them get their first clients and figure out how they can be pivoting or scaling. I’m sure there have been listeners who think, that sounds incredible. I’d love to be a part of Merantix. I’d love to be a founder and have my own AI startup, either as a technical expert, like a data scientist or a software developer, or as a more business oriented person. And they might even want to be doing that at the Berlin AI campus. So when you’re admitting people into the Merantix program as a founder, how does that work? How can they apply? And then what is the process like?
Nicole Büttner: 46:30
I would encourage everybody, obviously, to come and check it out. So basically what we’re looking for is, or let’s say, what is the problem we’re trying to solve as a venture studio? We believe if that to build a successful AI company, it’s ultimately, you can Google this, a dual PhD problem, maybe even a triple PhD problem. So you need to bring together extreme domain expertise with extreme machine learning expertise with business savvy, let’s say those are the three poles, to really build impactful applications. And that’s what we are trying to do in the studio and what future will show how successfully we’ll continue doing this, but so far so good. And I think if you are interested in such a interdisciplinary way of working, and if you fundamentally believe, yes, that’s the way to do it, then already, that’s a good starting point for you.
Nicole Büttner: 47:21
And so we’re looking… the profiles of founders are quite different. It could be more CTO type founders that are very technical, but think, oh my God, all this business stuff, like somebody needs somebody to help me with that; I don’t have the right person I will do this with. That could be one founder. I would say, another profile of founders, like people who come from abroad. We have two American founders, for example, already in the studio but who want to do this out of Europe, but out of Berlin. And then depending on how your biography was, maybe your connections are also a little bit limited in Berlin or in Europe. And we have a very rich network and are well established, so we are happy to provide a home. And then there might be people who, and we also have those profiles of founders, are serial entrepreneurs, who’ve already founded one or two companies, sold or had their first entrepreneurial experience. But they don’t really know machine learning that well. So they don’t have the domain expertise in machine learning. But they have good business savvy. They know how to make a product, how to build a solution.
Nicole Büttner: 48:28
And I think those are maybe some of the archetypes and obviously open to any quirky personality with a brilliant mind, who’s curious and entrepreneurial in the end. But it’s a very intense journey. I mean, we like to put a lot of our resources at the disposal of founders and Merantix Labs brings in expertise in also building first prototypes and giving some expert advice depending on the products. We also have a deep network of industry contacts, so we can bring in touch with your first clients. And then ultimately we’re very plugged into the AI scene and investor base globally. And I think that’s just a very good mix. And I think what I find most humbling is that people are just nice. Honestly, it’s kind of a stupid thing to say, but I would still buy a used car from everybody who works here. I’m not sure you can say that about everybody you’ve met in your professional life, but I can generally say that. And I learned something from the people around me every day, and I think that’s the environment. If you love that, then you should definitely check it out.
Jon Krohn: 49:44
Awesome. So you’ve done an incredible job of selling what a great opportunity it could be for a broad range of people, whether they are serial entrepreneurs or looking for their first venture as a technical person or business person. And even if they’re not from a Germany or from Europe. But so how do they literally apply? They just go to the website and fill out a form or something?
Jon Krohn: 50:06
Womp womp. Regrettably, at this point, the podcasting platform we use inexplicably and without notification stopped recording Nicole. The fortunate news is that this bug only happened to the final four minutes of her footage. And I remember the broad strokes of all of her responses, so we’ll fill you in right now.
Jon Krohn: 50:25
In response to my question about how people can apply to be a founder at a Merantix company, Nicole indicated that you can visit the Berlin AI campus in person to get a sense of what the facility and people are like, if you happen to be in the area. If you’re not in the area, or if you just think it’s more convenient, you can simply fill out a web form on the Merantix website. And we’ve included a link to that web form in the show notes. You’re also welcome to reach out to her on LinkedIn. Speaking of reaching out to Nicole on LinkedIn, I definitely do recommend following her on that platform to stay up to date on all the amazing things she’s doing for the global AI entrepreneurship ecosystem and for AI consulting in particular.
Jon Krohn: 51:04
My final question for Nicole, and the only other one that was lost because of the corrupted media file, was the classic final question on the SuperDataScience podcast. That is, whether she has a book recommendation for us. She started off by joking that she doesn’t read nearly as quickly as she might like and that her husband apparently reads two books a day. I presume this was slightly sarcastic for effect. But that she highly recommends Enlightenment now by Steven Pinker. To wrap up the episode, here is what I said in reply to her.
Jon Krohn: 51:36
Yeah, I too am a very slow reader, and that is sitting on my shelf and I am obsessed with the idea of reading it. I’ve read several summary posts of it with great plots and I’ve actually done… So in addition to the guest episodes on Tuesdays, I do these five minute Friday episodes on Friday, and for several of those, I’ve used information that I’ve gleaned from content associated with this book, on this idea of how technology and law and good governance and science, how all of these things mean that at this time in history we’re living at, by far, the best time to ever be alive, COVID notwithstanding. And it’s so easy to lose sight of that. So I try to bring that back into the picture with five minute Fridays. And at some point, when I do actually get to the book, I’m sure we’ll provide even more fodder. So what a great recommendation. Thank you, Nicole.
Jon Krohn: 52:35
All right. Thank you so much for enlightening all of us on the program today. Thank you very much for being on the program, Nicole, and hopefully we’ll get a chance to have you on again someday.
Jon Krohn: 52:51
What an extraordinary communicator Nicole is. Remarkably, while filming today’s episode, every eloquent phrase that Nicole said was off the cuff, without a single pause to think and without a single retake. I was left feeling thoroughly energized and excited by spending the time recording with her.
Jon Krohn: 53:08
In today’s episode, Nicole filled us in on how we can spark AI innovation by understanding the business problem clearly, by getting started early, and by cultivating a safe innovation culture. She talked about how we can make use of large unlabeled data sets by training models with unsupervised learning approaches, and then fine tuning our model with a smaller set of labeled domain specific data. She talked about how automation and reusable software components enable us to tackle separate but related AI projects efficiently, how the AI canvas approach developed by Merantix Labs, in conjunction with Nicole’s Alma Mater of St. Gallen, provides a blueprint for successful AI transformation within organizations of any size and any stage of development. Then she talked about how ideal AI startup founding teams consist of a triangle of someone with ML expertise, someone with domain expertise, and someone with business savvy.
Jon Krohn: 54:02
As always, you can get all the show notes, including the transcript for this episode, the video recording, any materials mentioned on the show, the URLs for Nicole’s LinkedIn profile, as well as my own social media profiles at www.SuperDataScience.com/543. That’s SuperDataScience.com/543. If you enjoyed this episode, I’d greatly appreciate it if you left a review on your favorite podcasting app or on the SuperDataScience YouTube channel. I also encourage you to let me know your thoughts on this episode directly by adding me on LinkedIn or Twitter, and then tagging me in a post about it. Your feedback is invaluable for helping us shape future episodes of the show.
Jon Krohn: 54:41
All right. Thank you to Ivana and Mario, Jaime, JP and Kirill on the SuperDataScience team for managing and producing another energizing episode for us today. Keep on rocking it out there, folks, and I’m looking forward to enjoying another round of the SuperDataScience podcast with you very soon.