Natalia Bielczyk 00:10 Hello, everyone. This is yet another episode of career talks by Welcome Solutions. And in these episodes, we talk with professionals who have interesting career paths and who are willing to share their life hacks with us. Today, I have a great pleasure to introduce Justin Shenk. Justin is the CTO and co-founder of VisioLab, an AI software company. He is completing an external PhD with Radboud University in Nijmegen. 

Natalia Bielczyk 00:42 And his background includes neuroscience research and building deep learning applications for Intel, and Peltarion as a Data Scientist. He is currently based in Berlin, Germany.

Natalia Bielczyk 00:53 Thank you so much, Justin, for joining us today. Happy to see you again, after some time. And I’m very curious about your career story from the early beginnings. Because I know that, you have quite a non-linear career path when you first started working in industry, then decided to do your PhD. And now you’re back in industry, this time with your own company. I’m very curious to hear your story from your own perspective.

Justin Shenk 01:25 First off, thank you, Natalia, I’m glad to be here. And I’ve been a quite a fan of your videos that you showed on YouTube so far. It’s really inspiring to see how you can bring together people who have similar backgrounds or different backgrounds. But bring them together on the same topic of how to figure out what you want to do, especially coming from a scientific background. It’s really an honor to be here with you.

Natalia Bielczyk 01:52 Thank you so much. Thank you. Could you tell us a little bit more about how you first started your professional career some years ago, when you found your first job after Master’s studies? Could you tell us a little bit more about your career decisions ever since?

Justin Shenk 02:14 Sure. Going back, it does feel like a long time ago when I started in biology. I started in research, and was doing neuroimaging research, looking at activation of the brain and patterns of the brain for speech and motor control. And going from research to industry took me quite a while. I think I was not sure what I wanted to do coming out of my Master’s degree in biology.

Justin Shenk 02:48 And there were many options in terms of, within research, there is a lot of different options. And then, also you hear stories of people going into industry and finding things that are interesting there. And being based in the science, but also having a lot of interest in technology, I explored some computational approaches. And that led me to returning to studies some years later.

Justin Shenk 03:16 I lived in various countries for some years before and had a chance to explore different ideas about what kind of career would be right for me. But it was feasible. I found this program in Germany called cognitive science, which was bridging from biology to mathematics, statistics, machine learning. And that gave me a chance to stay in the area that I was already comfortable with, which was understanding the really complex problems of like, ‘How does the brain work? How do biological systems work.’ Really, really complex things.

Justin Shenk 03:57 And then a framework that is much more straightforward, in a way, something you can model or write down on paper, like modeling systems. That was a great bridge for me, seeing something familiar, but also try something new. That’s generally a theme that I’ll probably promote quite a bit, how important it is to try new things. Especially if you’re coming from scientific research.

Natalia Bielczyk 04:25 Could you tell us a little bit more about the PhD program that you chose? Because from what I understand, it’s not like a classic way of doing a PhD. You did external PhD, which is in the Netherlands, not a standard. The standard way is to get a full-time four-year contract with the university and get the promoter and get your seat behind the desk. And become a member of graduate school, where you’re like in contract with maybe 50-100 other people who are kind of in the same PhD candidate cloud.

Natalia Bielczyk 05:06 And you have common activities and excursions and also seminars; This is more like a little community. But of course, you’re also packed to your own lab. You have a promoter, and you have lab meeting, and you have, of course, you’re a boss of your own project. You have to lead your PhD project for four years and end up hopefully, with a PhD thesis. But in your case, it looked a little bit different. Can you tell us a little bit more about how your PhD looked like and why you chose this external PhD?

Justin Shenk 05:45 Sure. My experience was a little bit different. For one, because I was starting much later in my life, and had already quite a bit of research experience behind me. The story goes that I was in Peltarion; working with Peltarion in Stockholm, Sweden. And on the board of directors is a computer science professor, Jan Bosch. I made clear to the company that I was interested in doing a PhD.

Justin Shenk 06:20 And he was very helpful and helping me find a supervisor in the Netherlands who I could do an external PhD with, which would allow me to do the PhD alongside my work. The topic of studies then, is like a bridge between the programming, which at the time was really my main interest to get more in-depth into programming and data analysis. And also, my previous background with biology.

Justin Shenk 06:48 Looking at the effect of diet on neuronal pathology. It’s quite a nice bridge, fortunately. And I was quite lucky to find a supervisor, Amanda Killian, who had enough data that I could work with. In my first conversation with her, the question was like, ‘Hey, do you have data? Like, how much data do you have? What kind of data?’

Justin Shenk 07:17 Because it was something where I knew I could make a contribution, and then also wouldn’t require me to be there physically in person to do experiments. So far, it’s been a really good relationship. And I’ve been really lucky to be able to combine these two sides of my life that I care a lot about.

Natalia Bielczyk 07:38 Cool. How did it look like in practice? How did it look like? Did you live in Nijmegen during your PhD and did you have a day job? Were you like, funded by the university? Could you tell us a little bit more what external PhD means?

Justin Shenk 07:55 Yes, sure. In this case, for the first year in Stockholm, based in Stockholm, and working say, 20% of my work week was on the thesis and a lot of my free time. that was working pretty well. But like, you also get the feeling that you know, at some point, it was also really quite stressful, because you don’t have much time left over for non-research things.

Justin Shenk 08:28 But I understand that I can least sympathize with a lot of people who have focused purely on research during their PhDs and how it really becomes like a camaraderie of people who are over-committed to research. I think I can appreciate that aspect of it. coming from then Sweden, back to Germany, and continued doing the PhD work.

Justin Shenk 08:54 At various times of spending more or less time on it. But basically, it was a lot of work. My schedule was not saying for most of the past three years, and then starting a company, a VisioLab last year. And it’s been a really good experience. I’ve been able to apply what I had learned and the skills I had before starting the PhD to the topic.

Justin Shenk 08:54 And I can only recommend, if you’re not sure about doing a PhD, then gain some experience or a specific skill that will be really useful for a team. And then when you join, you’ll be able to apply that without as much friction, and you’ll be more confident about what you can do with your skills when you start the PhD.

Natalia Bielczyk 09:42 Did you ever have like second thoughts? For instance, did you ever think to yourself, ‘Well, life would be easier for me if I chose to do a PhD, the regular way’.? You know, so that I only have one thing to care about which is my PhD full time. Or if I never really took a decision to do a PhD. Do you think this is, in hindsight, this was the right decision to do it exactly this way?

Justin Shenk 10:13 Interesting question. I’ve had a chance to meet some people the last few years who really inspired me about what a PhD can mean, like personally. Some people who just found a project and a topic that they love and want to spend their time doing, but would not be able to do this in industry. And if they did it as a hobby, they wouldn’t get paid for it at all.

Justin Shenk 10:38 That’s a category of topics and a mesh of people in topics that is, to me, like the ultimate reason to do a PhD. Because you just really love something, it’s a hobby that you can turn into a profession. For example, researcher, PhD student in Osnabrück, Viviana Kakerbeck, Clay as her married last name. She’s doing a topic of like deep learning with simulations where you have characters running around in arenas.

Justin Shenk 11:10 And you can do deep reinforcement learning to understand the decisions they’re making, based on how they perceive things. Super interesting topic, that would be really hard to justify to any kind of company like, ‘Hey, we’re going to do this really experimental thing’. I find it super inspiring. And makes sense to me if you’re going to do a PhD, to find something that you really enjoy doing.

Justin Shenk 11:37 And it’s often hard to know, before you start, what kind of modality you’re going to spend most of your time doing. Unless, you’re purely in the math direction, or purely in the biology, molecular biology direction. I think finding a niche where you can do something that you enjoy, it might be worth spending a couple years to figure that out before you commit to multiple years of a specific direction.

Natalia Bielczyk 12:02 But I also think that, it’s often the case that, projects are packed in certain ways. Like people who come to neuroscience. For instance, they often think that, neuroscience is all about thinking about the brain and cognition, and just thinking about how we reason and how emotions work. And in some ways, it is about, but the daily life it’s not associated with as much with philosophical discussions.

Natalia Bielczyk 12:38 You have a project, and most of it is just being down to the gallows and calculating staff or testing participants or building some infrastructure; it’s hard work. That is nowhere near to just deliberating about the humanity or the nature of human brain. The daily life looks very different from what you could imagine before you started a PhD; solely on the basis of the job description.

Natalia Bielczyk 13:06 I studied physics before. Studies in physics were kind of the same for me. In high school, I was interested in black holes, how the universe started and all these theories behind the Big Bang Theory, and theories behind how universe is structured and how it all began. And then I came to university and I realized that studies in physics are all about calculating tensors and all these complex formulas. And it’s just solving formulas.

Natalia Bielczyk 13:43 Nobody’s talking about where it all began, you know. There’s no such thing. Maybe once a year, at some yearly drinks with your professor, you can casually mention something and start deliberating for five minutes, but that’s it. And the rest of the year, it’s pretty much, it’s a grind, and trying to get skilled in solving formula.

Natalia Bielczyk 14:08 I think, it often becomes hard to, basically, figure out from the outside, how it all looks like. And you have to unfortunately, test it for yourself. But that’s also why it’s so important to change your mind sometimes and just, you know, ask yourself, ‘Is this what I really came here for?’ Because I think we often don’t do it early enough. And if I think about my own PhD, also, it wasn’t really what I expected.

Natalia Bielczyk 14:42 But I never really found time and energy and guts to tell myself, ‘Okay, this is maybe not what I should be doing right now with my time.’ But anyways, I’m happy that it worked for you and that you’re finally happy about your decision. I would like to hear a little bit more about your company, because I know that it’s going well for you. So, congratulations.

Natalia Bielczyk 15:08 Today, could you tell us a bit more about how VisioLab has started? And about how you found your co-founder? Where did you get your concept from? Why did you decide to work on this particular topic? And first of all, what the company does? That’s question one.

Justin Shenk 15:28 Sure. First, I’d like to mention on your previous point about how going into science and finding that it’s less maybe about those really broad questions and it gets quite specific. Reminds me of the saying that, ‘Going into science to explore the wonders of nature is like becoming a priest to meet girls.’ Like, you can do it, but it’s not the main point.

Justin Shenk 15:57 Science, like the romantic part of science, the part that’s really interesting and is quite separate from the modality of most day-to-day work. But you do probably meet interesting people that have interesting conversations. That’s true. And then to answer your question about VisioLab.

Justin Shenk 16:19 The story begins about a year and a half ago, where I was introduced to a CEO of a Jewish Hospital, where they were looking to do food recognition as an app for recognizing what kind of food and calories were being consumed and the volume of the food as it was changing. And I was in the position of estimating what would be the feasibility of this type of project, given a certain budget.

Justin Shenk 16:54 And I looked at it and thought about different ways to do it. And around the time when I was figuring out what would be the feasibility, and if I would be a good person to build this. I came across my co-founder Tim Niekamp, who was in some ways, he was quite further in the process. Because he had set up the company in Germany and he had already found interns or found people or students who could work on this project and get a cool prototype up.

Justin Shenk 17:23 We were coming from quite different perspectives. He’s coming from a business perspective, building a product, and I was coming from a more technical consulting perspective. And the timing was perfect, in a way because we got to leverage each other’s strengths and to build a team. Since last year, it’s been a rocket ride of changes.

Justin Shenk 17:48 I’m super happy that I’ve had this experience so far, and I see a lot of potential. The main challenge that we’re trying to solve now is how to use computer vision and AI, and machine learning, to recognize the food on trays for the purpose of automated checkouts. This scenario is that, if you go to a canteen or cafeteria and you put your tray of food under an iPad, how fast can we make the checkouts so that it’s automated, and it does the task of calculating the cost as fast as possible so that you can have an enjoyable meal.

Justin Shenk 18:29 And the benefit for that at canteens is that you can set multiple of these up at once. You can have 4 or 5 or 6 different checkout iPads systems and parallelized operations. That’s pretty interesting on the business side. And on the tech side, there’s so many challenges that you come across that are at least interesting to learn about; sometimes difficult to solve. But you don’t have a shortage of interesting conversations about what’s the best approach. That’s the fun part of it and the engineering side.

Natalia Bielczyk 19:08 I’m tempted to ask you a question. Since you know, I think everyone has a element of a con artists in them, even if you’re a good person. I think everyone has that button somewhere in their head. The switching that button on, I have to ask you. What if I put like one banana over another one, let’s say and just put the ashtray behind your sensor. What do you do then? Can you recognize that this is a fraud or not?

Justin Shenk 19:42 That’s a funny question. We hear that a lot. Also, for what it’s worth, I’ve noticed more business people asked this question than tech people. I don’t know if that indicates that you’ve shifted over already into the business side.

Natalia Bielczyk 19:57 I was screwed over enough times, you know.

Justin Shenk 20:00 That could be it. Like you’re saying, maybe scientists are just too sheltered. This is an interesting question. For sure, we can recognize with our system, some obvious types of fraud. But it’s not intended, in the current form, to be kind of to catch everything. Like if you think in the spectrum of, let’s build a whole store that has 100 different sensors, and does facial recognition. And now probably, it’s the store of tomorrow in China and the store of tomorrow, tomorrow, tomorrow in Germany.

Justin Shenk 20:37 But that kind of stuff is quite far away. What we’re talking about more is like, okay, let’s say you want to quickly scale up your operations, and you still have a cashier there who can keep an eye on things. Or let’s say that the fraud is the minimal case, because it’s an employee canteen, where your boss is looking over your shoulder anyways.

Justin Shenk 20:58 We’re starting with the kind of easy markets to go into. And from there, we’ll be able to figure out exactly where is the proper balance between adding more sensors versus making this super affordable and scalable.

Natalia Bielczyk 21:18 The answer is I can cheat.

Justin Shenk 21:22 Yes, if you want a list of canteens where you can cheat at. I can provide you with that afterwards.

Natalia Bielczyk 21:28 All right, good. You know, there’s like so much interesting information I get out of shooting these episodes, that is just, you know, beyond the original purpose. Okay, well noted.

Justin Shenk 21:43 Side benefits.

Natalia Bielczyk 21:45 Okay. Can you say a little bit about your methods? What type of software you’re building? Is there any, like proprietary technology that you’re building or you’re optimizing the current ways of …? I guess it’s a classification problem, what you’re solving. The visual recognition you do, like, correct me if I’m wrong. I guess that’s the classification problem. Are you optimizing the currently popular algorithms for that? Or are you building something new?

Justin Shenk 22:24 No, good question. We try a lot of different things. And especially in the beginning, we tried so many different approaches, from state of the art to YOLO, out of the box implementations. And we found that it makes sense. First off, that like, something that’s written for a paper is designed to work in an academic context. And then the exception is some things that are more clearly … For example, like a lot of things that Google research does, or Facebook research or, like bigger companies that publish research tends to have applications beyond academic settings.

Justin Shenk 23:08 It turns out that we’re currently using a combination of approaches. We do use object detectors like YOLO, and Faster R-CNN. But also, other unsupervised learning techniques. It turns out that there’s a lot of stuff that you can do with the information, when you take an image and you put it through one of these object detection models. Then the output of that is something, a feature vector, that you can use for further classification and get further information from.

Justin Shenk 23:39 And we’re currently looking into more experimental approaches to combine information that we get, like metadata from the dish. For example, if we know that this dish is a pasta or a salad, or a drink, then we can use the information from that text embedding to make additional classifications or to improve the performance of our model.

Justin Shenk 24:06 It does end up being more complex than something that you get out of the box. But I would say, like, the most complex stuff that we’ve seen is really like, … If you look at research papers that say state of the art, view shot learning, object detection. A lot of that is really niche stuff, and is not necessarily the thing that’s going to work in most cases.

Natalia Bielczyk 24:31 I’ve seen some research recently about comparing the results between two cases one where you work on the quality of the input data, and another way you work on the quality of the algorithm. And it was clear that the algorithms are usually good enough. It’s more about the quality of the data; that’s a much better thing to focus on.

Justin Shenk 24:56 For sure, that’s a good point. And that’s one thing we’ve also learned is that being able to control the environment where the algorithm is happening, where the data is fed into the to the model is the most crucial aspects. Being able to constrain the problem as much as you can really, really helps. That you can control for, for example, there’s lights overhead. They’re going to be flickering on and off, then that’s a pretty clear target to start with.

Natalia Bielczyk 24:56 In fact, that’s also how a lot of people who work in machine learning, consume or lose most of their time. They just really try to tweak the algorithms to get this wide 0.1%, you know, better success rate. But in fact, what they lose is most of their reliability on, is the data.

Justin Shenk 25:56 And then from that point of view, you have probably as much challenge of trying to think about, how do I get the user, like the user experience to be constrained as you do about how do I get the 0.1% improvement from a different model.

Natalia Bielczyk 26:19 I know that in this area of visual recognition, now there is a hype for deep neural networks. Are some of your methods, …? do they include deep nets, or not really?

Justin Shenk 26:36 Yeah, definitely. From the beginning, we tried a lot of deep network applications. And starting, as far as I know, this is standard for both of our competitors. And also, for research, for the state-of-the-art research, deep learning approach is really crucial. The advantage comes from the ability to take an image and to find patterns and images.

Justin Shenk 27:06 And for that convolution neural networks are the standard. That’s, of course, changed a lot in the last 10 years. It wasn’t the standard 10 years ago, and it’s the standard today. Then you can ask, ‘Okay, what will be the standard 10 years from now? I’m really curious to see where things would go.” It seems like things are moving at such a fast speed, in terms of technology.

Justin Shenk 27:30 I think it’s exciting to know that as a company in this space, like we have to be at the front of the technology. It’s one of the parts I like about my job is that I have to be able to anticipate the changes that will come in terms of, if it becomes, you know, even easier to recognize items. Say, one of the big tech companies releases a machine learning model that can automatically recognize anything as anything.

Justin Shenk 27:58 This is kind of, in a way, a worst-case scenario, and then make that free. Then you would ask questions more on the business side, ‘Okay, what is our unique selling point in that case?’ I can’t take for granted that what works today will be useful 10 years from now. But like you mentioned, you want to stay agile, and you want to be prepared to handle these changes as they come up.

Natalia Bielczyk 28:26 I fully agree with you. I think this is one problem with IT companies that it’s quite very competitive. And it’s often the case that the biggest player wins the market. As soon as someone is big enough to start charging, … to start making income from marketing and from the traffic, organic traffic, then they make the software free. For all the small companies that have a paywall, then this is like a disaster.

Natalia Bielczyk 29:05 That would be definitely my biggest fear if I was in that space. But well, let’s hope for the good. Maybe if you, indeed, specialize in some specific area and have better quality detection in that particular area. You know, maybe there is still space for small companies. I don’t consider myself an expert in this area, I don’t know, but let’s hope for the good. You recently found an investor which is very exciting. From what I know, you got some seed investment coming to the company. Can you tell us a little bit about that?

Justin Shenk 29:51 Yeah, sure. I think the press release will probably come after the airing of this video. But we’re excited that we’ve been able to secure our first large investor for the seed round. And that lets us focus on the next couple of years of development and growth. And the hustle has already been on for quite a while, it feels like. But in a way, it kind of legitimizes the work that we put into things.

Justin Shenk 30:22 And also helps me to shield the tech team from having to think about, ‘Okay, how long can we plan ahead?’, so now it’s much clear. And that’s brought mostly, at least at the moment, I’m seeing most of the positives of this. What challenges come from, taking an investment for example. It’s more likely, from what I understand, it comes with later larger investments where you have to make harder decisions about what path to go down. Because the stakes are much higher for everyone.

Justin Shenk 30:23 At this stage, I’m really glad that we can focus and that our hard work has been recognized. And that our investors see how the industry is going to change and how we’re in a good position to deliver that change.

Natalia Bielczyk 31:21 Great, so congratulations on your investment. I have more general question about this industry. What is actually possible with for AI algorithms in terms of visual detection today? I’m curious, because I’m wondering about that, regarding one of my projects. It’s unrelated to my company. But I’m thinking about whether or not it’s possible to create some tasks that are based on visual detection that are easy for humans, but very difficult for machines.

Natalia Bielczyk 32:02 And I know that Google CAPTCHA and like these authenticators online kind of use those tasks. Because, you know, sometimes you have to choose a set of pictures that represent certain category or so. Apparently, those are good tasks to tell the difference between a human and a bot. But I’m wondering what is or will be possible within the next 5 to 10 years with AI? And if there will be some point in time when AI is as good as human with visual detection? Do you think that it will ever happen?

Justin Shenk 32:40 Good question. As far as I know, the AI has beat humans on many, if not most visual tasks. I’m trying to think of the exceptions, like the captcha is a good exception. Not because they can’t recognize bicycles; because I mean, for example, it says, ‘Click the bicycles’. That AI can recognize bicycles without problem given enough data.

Justin Shenk 33:09 But the captchas where, for example, imagine you have an adversarial attack. You throw enough noise onto the image in a way that is designed to make it hard for the algorithm to recognize the item or to falsely classify it; those kinds of cases. That’s the edge case that I’m most familiar with in terms of breaking the AI or where humans are better.

Justin Shenk 33:41 And this is a rather special case, because it’s specifically designed for that purpose. If the question is, like, how will AI improve over humans in the future? I think as long as there’s data available, the AI could reach human performance generally. And if not, then it has to be specifically targeted to not perform well through adversarial attacks.

Justin Shenk 34:12 But I mean, that’s a really generalization. And then third generalization I might make on this is, there’s things where you don’t have a clear category to assign things. For example, like what makes something distinct or not. A person can recognize a dog after seeing a dog once that it’s a distinct thing; there’s something distinct about a dog.

Justin Shenk 34:39 And to generalize that concept, and then see a cat and to know that cat is a different thing. Teaching an algorithm or an AI system to identify things as different things then it becomes a bit of a philosophical discussion. Like, how is it children can learn a word once and be able to use that word appropriately forever afterwards. That’s incredible.

Justin Shenk 35:10 And I don’t know how do we measure the success of AI to do the same thing. Unless we restrict the task so much to say, like, ‘Okay, given labeled data X, can you predict Y’, that’s the a very supervised learning case. And the more ambitious machine learning approaches, or the unsupervised learning case, where it’s harder to define what the actual task is.

Justin Shenk 35:36 You can identify, ‘Okay, this word was used somewhat correctly, in the correct context, and somewhat incorrectly in the other context’. And then we’re getting to like actual human fuzziness of categories. In a way, like, technologically, I’m quite optimistic about unsupervised learning approaches. Because you don’t assume to have a black or white label for things that’s more accurate to how people think. But it raises a lot of philosophical questions, or just general curiosity about how to compare human intelligence with artificial intelligence.

Natalia Bielczyk 36:19 All right, very interesting. Hard to predict, for sure. I also know that while the algorithms are currently very bad with is sense of humor. Especially understanding irony, is almost impossible. And there are some trials to create algorithms that imitates human sense of humor, and generate jokes. And these are usually, you know, very bad jokes. Bad in a really bad way. Not bad so that they are good, but really just a completely, you know, misses the point.

Natalia Bielczyk 37:06 That’s also I think, good news. In the sense that, if you’re a person who has a sense of humor, then you should nourish and nurture, like, you should nurture your talent to generate humor. Because it’s not only something that can make people around you happier at work, and really leverages your value as a employee as a co-worker.

Natalia Bielczyk 37:37 But it’s also something that is very hard to replace by a machine unlike many other tasks. This is something people usually don’t take into account, that it might not be their core competence, but it might be. Probably in five to 10 years, it will be even more important, and more like distinct than now.

Justin Shenk 37:59 Relationships are still super important.

Natalia Bielczyk 38:03 They will be always and that’s one thing that doesn’t change. And human psychology doesn’t change, you can see it from the current Bull Run, is the same as in 2017. You know, people don’t learn; like emotions never change, really. And the basics of like how people make decisions and behave. It will never change, I think, even when it’s perfect.

Natalia Bielczyk 38:29 You know, if AI get to the point where it takes most of our jobs, and we don’t even need to work anymore. I still think human nature will not change so quickly. Anyways, sorry. I’m getting to, like, outer space with my thoughts. Next question I have for you is about possible future directions for your company. Since visual recognition and detection is like, as you said, very broad topic, and there are so many potential applications.

Natalia Bielczyk 39:10 Do you also think about your next steps? Or is that what you doing right now? So creating solutions for, horica, industry. Are you planning to set the scope of the company on this particular application? Or are you already thinking of potential future directions?

Justin Shenk 39:36 It definitely makes sense to get some traction before figuring out new directions. And at the same time, it’s useful to have in the back of your mind, what could be another way to go. For example, I think most of the people in the catering space have had to question some assumptions the past year, since a A lot of canteens have been closed due to the COVID. What I think is important, generally for us, is to set the scope for the technology so that it’s useful beyond the specific business case or use case.

Justin Shenk 40:17 Because the technology is there to solve problems to make people’s lives easier. And I think you can learn a lot from getting a product to market and to see how the reception is for specific products. And you can learn a lot about the user experience. And you can learn a lot about how your back-end technology is stable for certain situations. All those things you can take with you and put in another use case in other environments and other business case.

Justin Shenk 40:52 I think it is partly my responsibility to make sure that we balance the focus on the near term of getting to market and really understanding the problems that our customers are having, and how to add value to them. With building sort of scalable infrastructure and software that has applications beyond one specific sort of dependency or one specific use case.

Natalia Bielczyk 41:21 Also, one exciting thing about business development is that you don’t know the future. It took me a while to start treating this fact as positive rather than as a caveat. Initially, I was, you know … Because in academia, we like to think about our goals in the distant future as like, very well-defined goals. If you enter grad school, like most people still have this mindset that they want to land a tenure track position.

Natalia Bielczyk 41:54 And once you start building a company, you have to change this thinking and become much more flexible with your goals. Even the goal of developing a business might lead you, it sounds like a well-defined goal. But in fact, you can develop towards like 20 different directions. It’s not as well defined as it looks at first. I think it took me a while as well to change my attitude and approach.

Natalia Bielczyk 42:32 And stop thinking in terms of what precisely I want my career to be like in 20 years. But more like acting in this like action-reaction type of way, where you create products to solve problems. And if some problem doesn’t seem to be important anymore, or like you already created the solution or the solution is not even necessary anymore, then you have to change the scope; and this is fine.

Natalia Bielczyk 43:03 And accepting that this is fine, and you have to go forward. Now, I feel comfortable about it, but it took a while. But once you accept it, it gives a lot of freedom. And making decisions upon the rules you have, and the general guidelines and heuristics that you worked out. Rather than on the basis of some, you know, elusive goal that is on your mind, and that you’re planning to get to in 20 years, is probably just more enjoyable.

Natalia Bielczyk 43:43 And also leads to better career decisions as well. And that’s also what I’m working on at the moment, how to create those heuristics in the right way to open as many doors as possible. And make sure that you still focus enough to do your core projects, and develop them to the best of your abilities and how to keep this balance that’s really difficult.

Natalia Bielczyk 44:10 Can you maybe tell us a little bit more about the way you look at your career so far? And so, you were working in industry, then you did your PhD. You’re still about to graduate, but it’s kind of set already. We can say you’re a PhD to be, right now. And now you’re a CTO of your own company. Is there anything in your career path that you would do differently if you had the chance? Do you think about some of your career decisions as unnecessary or even harmful in a way when you look at it in hindsight?

Justin Shenk 44:55 Good question. I think, it’s hard to find in a place where I would definitely have done something different. I think in hindsight, you can say, ‘Well, if I knew then what I know now, then I would have started sooner on the same path’. But I can also pretty easily say that one of the reasons why I really liked the path I’m on now is because I tried out the things that I didn’t like as much.

Justin Shenk 45:23 It’s hard to know that I would be satisfied unless I had made those experiences. I think that one thing that I would recommend to say younger version of myself is to just be a little more proactive and seeking out new experiences, contact people more, network aggressively if you have to, be persistent. I think these are skills that aren’t taught directly to you in a scientific education, but end up being super, super valuable.

Justin Shenk 45:56 And I think it’s kind of like, it makes sense that on the spectrum or like the skill wheel. Like in science, you can learn analytical skills really well. And on the other end of that spectrum, you’ve got maybe more social skills. And it just turns out, from my experience, that social skills are so much more valuable than given credit for in a scientific education.

Justin Shenk 46:25 I’m not sure how you found it with physics. Like in biology and chemistry, or physics labs, we did have interactions, we had teams. But I think the skills that help you to manage a team of 10 or 20 people are probably not going to be developed sufficiently in the case of pure academic laboratory work to say, undergraduate level.

Justin Shenk 46:53 Taking leadership roles, if you can, if it suits your personality. It’s things that I found really useful. And not being afraid to message people across the internet and see if you get a response back. And see if there’s some way that you can help other people in a way that also helps yourself. Volunteer if you have the privilege to do so and get experience that opens your world a bit.

Natalia Bielczyk 47:25 And I also think that when we network, we often network in a way that we ask for things. When you think, ‘Okay, I need this’. Let’s say, I’m looking for a job, I need a job. Who can provide that job for me, or who can provide information that will help me get a job? And we start interacting with people in a way that we want something from them.

Natalia Bielczyk 47:50 We are just searching through our network and trying to shake out something out of our personal contacts in a way. And what works better often is doing exactly the opposite is called Net Weaving. It’s thinking about, what can I contribute; what are these different little bits that I can do. For instance, if I know that someone has a problem, and someone else has a solution. These people don’t know each other, they come from my different circles.

Natalia Bielczyk 48:20 One of them is my high school friend. The other one is like a person I just met at a business meetup. But I can make them contact to each other and maybe it works. But it’s just this heuristic way to build networks, but in a different way; thinking exactly the opposite. And thinking about how I can contribute to other people’s careers and lives.

Natalia Bielczyk 48:47 For a person who is doing that, it usually brings overall more benefit than networking. Because, you know, this is also very human, that we feel we want to contribute and give back. If we get help, we also want to help back. People who have that approach usually benefit more from the time they spend on building networks, than people who have that attitude that they have a goal, and they know what they want to get out of the network.

Natalia Bielczyk 49:26 I’m saying this, because it’s also often the case that as PhDs, we have a lot of this career advice from everywhere. And this topic of networking is being repeated every time. It’s like one of the most popular pieces of advice you can get; network, network, network. But you have to also know how to network. You know, as with everything you can do it in a way that will make you only lose time.

Natalia Bielczyk 49:56 And you have to also recognize relatively early on, if someone you talk with is just full of themselves or they really know what they’re talking about; or if you’re networking the right way. There are ways of spending an evening, let’s say IT meetup, that will largely increase the strength of your professional network.

Natalia Bielczyk 50:24 And there are ways to just lose time, and just spend this time on meaningless conversations in their own corner of the room of the people who you don’t have anything in common with, and you will get nothing out of it. To network properly, it requires much more skill than to just network.

Justin Shenk 50:47 That’s true. That’s a good point. You remind me that there’s no like silver bullet for happiness; like there’s no one solution. And also, what may be helpful for one person, a totally different approach might be helpful for someone else. The way you describe Net Weaving, it’s reminds me of how valuable is just to understand what other people are going through.

Justin Shenk 51:12 Like you say, if you’re just trying to find a job. If you’re just trying to optimize for one function, then sure, keep hammering at the same nails that you have. But if your goal is to figure out what do I want to actually do, then in that case, it might be more relevant to get a wider view of the kind of things people are experiencing so that you can figure out what place could you fit into.

Natalia Bielczyk 51:43 And sometimes it’s also counterintuitive thing to be more agnostic. This is also a problem that many people make; the mistake many people make during their career. They aren’t oriented on themselves. They don’t spend enough time to define better what they enjoy to do and what they want to do. And then, they wake up at the age of 30, or 35, having no idea whatsoever; or they never wake up.

Natalia Bielczyk 52:12 I see this is a counterintuitive fact. But I think the world would be better if we all were a bit more egoistic. And because there’s a lot of misery coming from the fact that we don’t think about ourselves early enough. And so, that may be a controversial thing to say. But I think if we spend more time thinking about ourselves, we would be also in better places, like doing more efficient projects in more efficient works and solving more problems for others.

Justin Shenk 52:47 It’s an interesting point. I haven’t heard an intellectually satisfying definition of self, to help me agree with that statement. For example, it’s not obvious to know the difference between me is an isolated individual, and my group of friends; my family, people I care about. I think the pure egoism doesn’t make sense, obviously.

Justin Shenk 53:12 Because like, I don’t know anyone who I care to admire who is purely egoistic, and has no network of concern outside of their individual self. But if you say, generally, make rational decisions or something. Generally, consider that, you know, don’t be overly idealistic over these kinds of things. If that’s what you mean, then it’s easier for me to agree with.

Natalia Bielczyk 53:45 I think we have to slowly come to the end of this episode. Last question I have for you is a standard question I often ask. Is there any general career advice that you would like to give to young people today who are thinking about the future? And who might potentially also think about their own business, but they don’t feel they have guts to do it? Is there anything you’d like to share with them?

Justin Shenk 54:14 Yeah, sure. I think all of the things that I would share, maybe seem obvious if you take it at face value. But generally, understand what motivates you or what guides you, if that makes sense. And to know that, if that you’re guided by a sense of fulfillment that comes more easily through setting out on your own venture. Rather than from, say, taking something that has been handed to you then, it’s useful to know.

Justin Shenk 54:58 Whether you’re a scientist or an entrepreneur. If you really get a thrill from being the first person to discover some idea, it’s good to know that; it’s good to be aware of that. And to realize To what extent that shapes your motivations. I think the rest is something that people will figure out on their own. But I think it’s just really hard to generalize on this, but it’s a good question.

Natalia Bielczyk 55:27 Great. Thank you so much, Justin, for being with us today and for sharing your story and your invaluable advice. Good luck with VisioLab. I’m looking forward to see more updates and let’s see how it all pans out. I’m positive, I think, you’re doing great. I’m curious what happens in the future. I’m curious about which restaurants I can. The full, full, full list.

Justin Shenk 56:00 Sounds good. Will think on that afterwards. Thanks a lot.

Natalia Bielczyk 56:03 Thank you so much. Thank you for joining and for all of you guys who came to the end of this episode. If you like this content, please subscribe to this channel. And please leave us any comments and questions you might have; we will get down to them and also every single one of them. Looking forward to this and take care. Have a good day, everyone.