Natalia Bielczyk 00:10 Hello, everyone. This is yet another episode of Career Talks by Welcome Solutions. In these meetings, we talk with professionals who developed interesting careers and are willing to share their life hacks and valuable tips on how to develop your career with us. And today I have a great pleasure to introduce Fabio Gori. 

Natalia Bielczyk 00:32 Dr. Fabio, first completed his Master’s degree in mathematics at the University of Pisa in Italy. At that point, he began researching the biomedical field and he liked it so much that he decided to keep working in experimental sciences. Therefore, he went for a PhD in Bioinformatics and Data Science at Radboud University Nijmegen.

Natalia Bielczyk 00:53 The focus of his research was developing computational methods or the analysis of microbial communities. After completing his doctorate, he became a Research Fellow at the University of Exeter in UK, where he was using quantitative methods to study antibiotic resistance. He later decided to move to the industry, and so he worked in the German biotech for precision oncology as a Bioinformatics Data Scientist.

Natalia Bielczyk 01:21 He moved back to the Netherlands to work as a Data Scientist at the AI startup, Machine2Learn, where he works both on industrial and biomedical projects. Welcome Fabio, thank you so much for being with us. And I would like to now hear your story from your own perspective.

Dr. Fabio Gori 01:39 Thank you, Natalia. Thank you for inviting me here. That’s a good summary of my story. I mean, I was studying mathematics in Italy and I really like mathematics. But when I started doing more like application, applied research, I really felt it was my call. I really felt like this is what I have to do. And I felt like much more, and I can say, excited about the thing I was doing.

Dr. Fabio Gori 02:02 Then I decided to do a PhD and decide to do it abroad. And I wanted to try more like a … Initially, in Italy was more about biomedical research. I wanted to do something a bit more biology and I ended up in Nijmegen to study this PhD on metagenomics. You’re basically dealing with DNA sequences, sampling from community bacteria. There were problems about finding what bacteria are there and seeing what was wrong when you acquire the data.

Dr. Fabio Gori 02:38 And I felt somehow lucky because it was affiliated with machine learning group. I fell in some way lucky because I realized that this DNA sequencing was something that was growing. I mean, at that time, the cost of DNA sequencing was dropping at the speed of Moore’s law. I remember exactly like Moore’s law to be honest, but like every two years, it was half the cost. But then even drop even faster. After a few years, it went even faster.

Dr. Fabio Gori 03:07 It was still okay; this is a good place to be working with DNA sequencing data. But at the same time, I was in the machine learning group. And I felt like, ‘Wow, this is powerful’. I mean, I already did a bit of machine learning directly in Italy. But I really felt it has potential. In some way, I had the perception that I will find a job after leaving academia.

Dr. Fabio Gori 03:26 But after my PhD, I mean, I like that. But I felt like in bioinformatics, that was so much like say ‘grunt work’. Like, work you had to code a lot. Sometime I felt like my mathematical mind was not used enough. And so, I decided to do a postdoc involving a bit more mathematics and that’s why I end up in Exeter.

Dr. Fabio Gori 03:55 Because traditionally, the United Kingdom is the best country for mathematical biology in Europe. And there it was a combination of bioinformatics & statistical analysis, but also like mathematical modeling of experiments for studying microbial resistance; that’s antibiotic resistance. This is a big problem that people are not talking enough about. It is already killing a lot of people unfortunately.

Dr. Fabio Gori 04:22 And then I said like okay maybe it’s better to I mean like I felt like moving from jumping from country to country to research is now my life. I want to stabilize more; wanted more stability. I don’t want to think, ‘Okay, where we going to be in two years?’ Like, in which country we want to be in two years. And so, I decided to move to the industry and end up being in Germany in the biotech German world.

Dr. Fabio Gori 04:48 I mean, I changed a couple of jobs. But still it was an interesting, the first one was, because it felt like both my PhD and my postdoc was among the technical skills acquired in these two jobs were both useful in this new position. I felt lucky. It was really like I tried a transition where my technical skills will be put at use. I was really happy about that.

Dr. Fabio Gori 05:14 I felt like I was using them for something. Interesting at the end that somehow, my mathematical background would turn out to be more valid in the industry than in academia. I guess because you don’t find many people in bioinformatics with a background in mathematics.

Dr. Fabio Gori 05:33 Whenever there was something come up kind of like complex, so like a certain level of mathematical knowledge. There are not too many but there are quite some types of problems, they’re always coming to me. I was using my mathematical skills more than when I was doing bioinformatician and that’s something I really like.

Dr. Fabio Gori 05:54 But then I said okay … I felt also at a certain point, I felt that Germany wasn’t exactly my place. And I always wanted to come back to the Netherlands and focus more on the artificial intelligence or real data science. Because in bioinformatics … I was working as a scientist in bioinformatics. But the data is so complex and difficult that you spend much more energy on the data preparation, and also the topics are so complex, that you really understand that you need domain knowledge and is much more important.

Dr. Fabio Gori 06:28 I really wanted to put more mathematical background at use, still in data science role and to learn more methods. And this job from this point of view was quite good. I would say perfect because since it wasn’t a startup, I could also like somehow develop the job to fit my skills. And since this is a consulting job, you have to work in different projects with different people. And therefore, you’re pushed out of your comfort zone and learn new methods, for example, on your own skills, new libraries for programming. That’s what it’s really like.

Natalia Bielczyk 07:06 Fantastic, thank you so much Fabio for sharing your story. If I may, start this conversation from asking you. When did this final decision happen? At which point in time, specifically you said to yourself, ‘Okay, this is it? This is when I have to take the decision to move to industry’. Do you remember that day?

Dr. Fabio Gori 07:32 It’s not exactly one day. Even at the end of my PhD I wasn’t sure about doing a postdoc. I was thinking because I said, ‘I like doing research.’ But I know that after a certain point, I mean, you stop doing research in academia. And also, I liked what I was doing was more as a job. I mean, it was more bioinformatics I’d say. At least my postdoc was, … I think in some way, I liked it more because it was more mathematical.

Dr. Fabio Gori 08:08 I mean, my PI at the time told me, ‘You can try a postdoc. But after a poster, you have to make a decision.’ Otherwise, when you do the transition it will be harder. I found this a very short postdoc, two years. From an academic point of view, I will advise people not to do two years postdoc unless it’s a really strong continuation with your PhD. That was not my case.

Dr. Fabio Gori 08:31 Because by the time you start getting ideas, you’re at the end of your postdoc. Basically, by moving your field, you’ll start getting ideas after one year and a half. And since academia is all about ideas, that is not a good time, of course. As I said, the group I was working with was a combination of people doing systems biology, mathematical modeling for biology, and experimental biology.

Dr. Fabio Gori 08:58 Nobody was doing bioinformatics; I could cover it and interface with them and do a bit of mathematical biology. It was fine, there was work for me already. But I could not be the one coming up with ideas at this stage. But anyway, in that case for me was fine. A big moment, I think it was after I did my PhD defense. It was like six months I think or seven months after I started my post doctorate.

Dr. Fabio Gori 09:25 Because I came back and I felt like my brain for a month refused to work. Because it felt like, ‘Okay, you made me do so much work and then what you get?’ You get your piece of paper; you have to keep working even more maybe. I felt like, maybe by working more, producing more in academia, you increase your likelihood of getting a job. But that doesn’t translate into money that much. I don’t know why.

Dr. Fabio Gori 09:54 I felt in my mind, I didn’t like this translation. And also, I felt this connection between seeing results directly. Of course, the things that we’re doing is very useful. I knew that at least in my case, one way or another, the things that we’re doing was very useful. Especially my post doctorate, it was antibiotic resistance. It’s a very important topic.

Dr. Fabio Gori 10:18 But on the other hand, you feel like, this idea that you have to make a paper then submit it and then waiting for that, you know. It’s like have this waiting-time, anxiety, hoping that it’s too rushed is the other reason. It’s not like you daily deliver and you get satisfaction daily. This delay in time also for me was quite stressful.

Dr. Fabio Gori 10:42 And I wanted something with a more directly impacting reality. Although what I did already had, but something really more up to the point. And I never looked back, honestly. I never looked back. I mean, there are things that I miss about academia. But overall, no. I don’t want to be back.

Natalia Bielczyk 11:01 Great. Referring to what you said about not doing research in academic career. I have to say that recently there was some study here in the Netherlands by the courier and colleagues from Rathenau Instituut. And they analyzed data from 16,000 PhD graduates that graduated within the last 22 years in the Netherlands.

Natalia Bielczyk 11:28 And they found out that vast majority of grad PhD graduates still has jobs, industry jobs, that involve research to a certain extent. It was anything between 60% and like 85%, depending on the field of study. And among those PhDs who hold these jobs, the mean amount of working time spent on research was also more than 50%.

Natalia Bielczyk 11:59 They not only have jobs after research, but they also really spend this time on researching stuff. Which is probably much more than if you integrate the research time across PIs in academia. Probably it’s much less than 50% research time.

Dr. Fabio Gori 12:19 That makes sense. Yes.

Natalia Bielczyk 12:21 In the end, the conclusion was that in industry researchers do more research than in academia.

Dr. Fabio Gori 12:28 Yes, it makes sense. Especially, if you’re a PI it depends a lot on the field. But imagine you’re doing biology … I mean, like for now, things are going to change or changed, I guess. Because even academia people are starting to use robots in lab. But still, the job of a PhD, a lot of time goes into the lab and when you become Professor, so this time doesn’t get reduced.

Dr. Fabio Gori 12:53 If you’re doing computational research, like I was doing, is a bit different. Because you can carry your laptop in an airplane and would still be working on your algorithm. As you said, it really depends on the field. But in general, I think you’re right. You do more actual research in the industry. Of course, it depends also on your career path. If you go for becoming a senior or C11 in a company. It depends also on the career path that you should take.

Dr. Fabio Gori 13:21 And I think to keep it in mind, so that’s a big issue. But it still depends on the field that when you are doing innovation, there is an obvious like “The Ghost of a Hobbs” event are becoming obsolete. Like, what you know would become obsolete. Depends a lot on the field, depends on what your knowledge somehow and trying to have a local … I mean, regarding mathematics, my mathematical knowledge it doesn’t get that much obsolete. That’s a good asset for studying mathematics.

Dr. Fabio Gori 13:50 But still, like the programming languages using now, they’re not the ones used 10 years ago, when I saw my PhD for sure. Although, like eight years ago maybe, but like the libraries has changed, so there is this continuous innovation. Somehow, you have to keep it updated, this takes energy and effort. It also depends on the field. I know in some companies, for example in IT, the standard path is that after a certain point, programmers become managers for example.

Natalia Bielczyk 14:21 Indeed. Just a quick question about programming then. Are there any other programming languages that are important for today’s data science then Python?

Dr. Fabio Gori 14:35 I mean, R is still important; depends a lot in the field what you’re working on. In biotech, it’s still very important because uses Bioconductor. These collections of libraries. The libraries are what matters essentially. Although, I honestly prefer Python than R.

Dr. Fabio Gori 14:54 But there is an emerging language that is Julia. I haven’t used it yet. But I’m really tempted on trying to move to that. Partially, for the speed. But also, because of what I heard about the way it structures the code. It has some features that are really, really helpful, that are better than object oriented, for example, or a pure functional setting.

Dr. Fabio Gori 15:17 That’s one language, but for now, Python is still staying. Again, the advantage of Python is that it’s somehow the best or the second choice in many fields. Whatever you’re doing, there is a library, which is quite okay. The reason why during my PhD, I started with MATLAB and then I moved to Python. Because in MATLAB, you couldn’t really do bioinformatics.

Dr. Fabio Gori 15:37 And in Python, although it was the beginning, you could definitely do. I mean, the radiator was NumPy, so this is a computational library. There was no real library for machine learning at the time. But still, it was much better than if we do work in MATLAB. And plus, there was really quite robust library for bioinformatics.

Dr. Fabio Gori 15:56 I really felt like that by switching to Python, thanks to the libraries it had, I could do everything I needed to do. That’s why I stayed there. As I said, I see Julia, I think could emerge like in five years would really be something.

Natalia Bielczyk 16:15 Okay, well, that’s good to know. You know, maybe some viewers who are watching this episode, maybe you will get inspired to check Julia, and check what it is. I personally, never tried Julia. I don’t feel like I have an opinion here. Okay, great. Could you tell us now a little bit more about the data science market at the moment?

Natalia Bielczyk 16:41 Because data science is like a very hot topic and it’s one of the areas where most PhDs look towards after; once they ponder options. Definitely, there are some groups of jobs or groups of techniques you might be specializing in. Or perhaps the data scientist is supposed to be able to analyze all kinds of datasets. I presume, not. Can you come up with a systematic analysis of how this data science market, job market, looks like?

Dr. Fabio Gori 17:26 I will say that the science market is evolving in three different directions. This is what I read in a post but it was convincing. Somehow like, these generic figures that people missed from academia is branching in three directions. One is the AI researcher. Somebody who’s really developing new algorithm, prototyping new algorithm.

Dr. Fabio Gori 17:50 The data engineer, that’s somebody who’s taking care about the deployment of the model and all these things revolve around the model the data. And then something in between the was like the AI engineer, so, somebody that maybe can doesn’t elicit is not is called the core business to make novel developing models.

Dr. Fabio Gori 18:15 But definitely understanding the can use of standard models and packages. And it also can bridge that to the deployment somehow; so, something in between. I will say that as the time goes by, these libraries are evolving. 10 years ago, you had to implement the algorithm yourself. And now, other libraries would already have it implemented.

Dr. Fabio Gori 18:38 And this technical part is becoming more and more important because more industries switching into AI. I would say the central figure, design researcher, I would expect to grow. My figure is quite old school, as a called data scientist. Because I think my core is more like in modeling the problem, finding the right formulations because my background is in mathematics and asking the right questions.

Dr. Fabio Gori 19:06 In this case, my position is more like, I would say, interface with the domain expert. The person talks to the main expert and try to formulate the problem in a good way. And then, there are people who could work downstream. And not realizing the last couple of months, but I was looking on how, reading about how the pandemic impacted the market.

Dr. Fabio Gori 19:34 It seemed, initially, that there was a reflection of job offers like in every field. But then they came back because we just adapted. And of course, my company naturally got a project out of Coronavirus. In some way, it’s even driving more automation and therefore also AI. Because usually, Artificial Intelligence is on top automation.

Dr. Fabio Gori 20:03 I mean, unless there are people working in a specific field that was impacted by the pandemic. I don’t know, like companies doing like tourism somehow. Like, the no air traveling. People doing data science for traveling, data science for a touristic website. Unless you’re really in this specific field that were heavily impacted. I would say that it’s still quite good.

Dr. Fabio Gori 20:38 Initially, some people said there was an increase in jobs in the banking and insurance sector. And they said this is consistent with what happened in 2008. There was the crash in the beginning, people had to find a way to save money. And therefore, they used analytics. It was not data science and math were analytical; people looking at the data.

Dr. Fabio Gori 21:01 I will say that we were quite okay. The big question for me, I will say is like in 5-10 years, what’s going to happen between two trends. straight like. I was thinking about how are we going to be or get obsolete. In some way, it might be that, for example, deep learning is increasing. Although, I also have some criticism about deep learning. And I know some deep learning, but I’m not known to do something different.

Dr. Fabio Gori 21:28 Maybe some technology will emerge, and that will be just obsolete. Or maybe there will be more people doing data science, and therefore it will be less demand for them, so the salary would be reduced. Or maybe that the demand is so high, there will be more automation and therefore, there would be less jobs. I mean, it will just need people, a small amount, to manage existing stuff. I think I’m more in favor of automation. I think it’s the thing that’s going to happen more likely.

Natalia Bielczyk 21:59 I share your concerns; I have to say. Because, indeed, data science as a service is quite expensive. Especially in countries like the Netherlands where labor costs is, in general, expensive. Some data science or machine learning engineering jobs, indeed require some specialistic knowledge and experience. But some of them boil down to linear regression and cleaning the data a bit.

Natalia Bielczyk 22:33 I think at least some percentage of these jobs can be automatized. And since it’s such a high cost, there is also good incentive to do so. I mean, I also worry now when I see how many PhD graduates take that road. Then I always wonder, at what point, will the market get saturated. Because there are just more and more people in the space and potentially, the demand might be dropping.

Natalia Bielczyk 23:09 Indeed, in 10 years, I cannot tell as well. But that’s something you cannot really predict. Because at the same time, there are more and more different types of machines and different types of algorithms. Who knows? Maybe there’ll be even more jobs in that space.

Dr. Fabio Gori 23:32 There was a friend who contacted me recently, asking that she wanted to jump to data science. And I told her, ‘Why don’t you become a project manager?’ I’m kind of discouraging people to make the jump. Because I think that there is this hype or maybe they underestimate the level of knowledge. For me also, it’s hard to say. For me, some stuff it looks easy, but because of my background maybe for them it’s not.

Dr. Fabio Gori 23:56 And I’ll say, if you really want to be a project management, this a more stable career and more like a liberal standard career; and is a common jump. But also, if you want a bit of science, I would say, look at the sciences. Go to the science where your domain of expertise is valid. Because in this way, like worst case, it will just be with interface between data science and then the domain. And naturally, you will be able to make a career out of that.

Dr. Fabio Gori 24:25 I will say, it can be very bad. For example, one of my friends in the end, he was a biologist but knew how to code. He was a Bioinformatician also. I mean, he could do bioinformatics, could understand AI. And now, he has quite an important position in a pharmaceutical company because of this knowledge. But his background is a biologist. For example, what I can see, especially if you want to go up in more like a higher-level, like an executive level. The domain knowledge can matter more than the knowledge of data science.

Dr. Fabio Gori 25:02 I would say, it’s a skill that you can acquire and of course, will be valuable. But then probably your safest spot, and also maybe the most available. Look at this spot is to be on the connection between the two; between the domain and the science. And more on the domain or like an executive of a company integrating the two.

Natalia Bielczyk 25:24 Very good, very good advice. Thank you for this Fabio. Could you also tell us a bit more about what you think are the similarities and differences between data science in academia and in industry?

Dr. Fabio Gori 25:40 I think the main differences that are in academia, although when academia was doing some stuff applied, but there is a lot of focus on developing new algorithms. You think of a problem, you fight for getting a good accuracy and stuff like that. In the industry, it’s all about the data and they really see like data as the new assets. But I can tell you from an antitrust point of view, data is the new asset.

Dr. Fabio Gori 26:12 It’s something valuable. Because if you have the right data, any team somehow in a certain amount of time could replicate what you have, if you get the same data. From that data, there’s not that much you can do. Data is really the valuable thing. And I can tell you especially, when I was working in the Indian biotech world, a lot of effort that goes into data preparation, so understanding if there is bias.

Dr. Fabio Gori 26:38 You have to be very critical, say paranoid about the data. Is the quantity enough? What are the levels of error? But especially, the bias. Try to see if there is a bias, try to think about possible bias and identify them. Because if the quantity is enough and you understand the bias, maybe you can still do something.

Dr. Fabio Gori 26:58 But when it is low and there is a bias against what you want to do, then there’s nothing you can do. It’s very data dependent. That’s the main thing, the focus is much more about the data, I would say. There you see that the domain knowledge matters. you have to really understand the problem, you have to understand what you want to solve. There is a lot of effort that goes into this translation; this formulation of the problem.

Dr. Fabio Gori 27:21 Talking about the domain expert, asking the right question. Tell them what is your viewing about the things to check. Also, check what they tell you because it happens quite, … I mean, in the minority case. But still happens that some of the understanding of the problem when you see it in the data, it doesn’t emerge.

Dr. Fabio Gori 27:41 There’s also often a misconception about the domain expert. Not in everything, but a minority of things that could be relevant. It’s more like having good understanding of the data and then you build up something. If you don’t have a good understanding of data, you really risk to build something that just predicts the obvious, or something that doesn’t generalize.

Dr. Fabio Gori 28:00 Unfortunately, sometimes the clients really want to push you to go to the model as soon as possible. But then they’re going to be … I mean, it’s sometimes tough to convince them. This is the data but also, I think that there are some questions that people never ask in academia. Never asked like, they don’t ask them much in academia, but they’re very relevant. For example, imagine you’re doing a binary classifier; something classified within two things.

Dr. Fabio Gori 28:23 I imagine, I don’t know, between healthy people and sick people. In real life, what you’re going to have is most of the classifiers in the literature, they’re taught for working on data that are balanced; or the number of sick people and healthy people is the same. But in real life it’s not; it’s very imbalanced. The sick people likely are a minority of the less few than the healthy people.

Dr. Fabio Gori 28:48 But also from a life prediction point of view. You have to ask yourself, probably like predicting somebody that is sick is much valuable than predicting something that somebody is healthy. But also, the type of mistake you make. If you say that somebody is sick while they’re healthy. You made this mistake, what is the cost of the mistake. This is hard to say.

Dr. Fabio Gori 29:12 Your domain expert cannot tell you probably. But at least they can tell you if it’s higher or lower than the opposite mistake. If you classify somebody that is healthy, as unhealthy, I mean, these two things. I mean, AI it’s obviously a mistake, and sometimes will do even badly. We have to say that, you know, often AI is not enough.

Dr. Fabio Gori 29:36 You have to combine with many computational methods together to compensate with the mistakes, detect these possible mistakes, and avoid a worst-case scenario. And also tune these models, so that it takes into account the predicted different values. And there are some errors that are more acceptable than other types of errors. And these are things that you don’t cover in academia, or usually don’t cover in academia.

Natalia Bielczyk 30:07 You have to be much more practical, I guess. What happens if you do a project for a client and nothing comes out of the project? In academia, we have the tendency to torture the data and just mold it as long as it takes, so that we have our desired p-values. Which is commonly referred to as p-hacking. Because the overarching goal is to get some statistically significant effect. But what do you do if you do a project in data science for a client and you clearly see that there is nothing in the data, what do you do?

Dr. Fabio Gori 30:44 The best thing is to have like two weeks or something like that. A few weeks of like agreed job, when you get paid and you have a first look at the data, and you get a bit of a feel for it. If this is enough or is not enough, then you never know. If I would believe more like any commercial, but I will say I cannot promise it’s going to work. But I can promise that it will give you useful insights about your data.

Dr. Fabio Gori 31:07 What I would do, especially what I was doing when I was working … The fact is that, I always had ‘clients’ or customers even in academia because I was working with doctors, with biologists. I was trying to ask them, what are the questions they wanted me to answer. And then, I was coming up with answers to the questions. Sometimes the answers were something they expect. Sometimes that they did not expect. Sometimes it was an error in the data but sometimes their view was wrong. And that was valuable for them.

Dr. Fabio Gori 31:37 My point is that, if you integrate your developing AI model with something that is like an exploratory data analysis, so having a better understanding of the data, you may come with information that is valuable for them and they’ll be fine with that. For example, a client told me he wanted to do this. We said, we think there is a huge amount of money that could be saved.

Dr. Fabio Gori 31:59 And we look at the data, it wasn’t actually. It wasn’t a failure really. I mean, it was a success if you think about it. Because saw the reality was different what they thought; it was going to happen for sure. Unless they’re like a very advanced company already doing some data analytics insight, nobody’s ever looked at the data. Having just a look at the data is a value for that. I think it’s important to keep expectation low and to say that at least you will focus on letting the data speak and answer the questions.

Natalia Bielczyk 32:35 Very good advice. If you communicate with the client from the very beginning and let them ask the questions then you provide the report for the questions they suggested. In either way, whether or not you have some interesting facts, I guess. It is what the clients asked exactly. Good to know, I was curious what happens.

Natalia Bielczyk 33:05 Because I thought you know, that if a client … I never worked in this type of job so I just don’t have a feel for what type of expectations do these clients have. But as long as you can manage expectations, I think it should be fine. My next question would be, how does your working week look like typically?

Dr. Fabio Gori 33:29 Before or now with the pandemic?

Natalia Bielczyk 33:32 Both maybe. Because we don’t know how much longer the pandemic will take, so maybe this is the new standard.

Dr. Fabio Gori 33:41 I recently relocated to a new house so I’m now finding a new normality. Before I mean, I was living in Amsterdam now I’m commuting from the north. And I mean before the pandemic let’s say, but this commuting didn’t last too long, like six months. And in that case, I was coming in the morning. We follow a bit like the agile structure. We have a stand-up. We give a brief update with what we did the day before, what they’re going to do, any blockers and then we start working on our projects.

Dr. Fabio Gori 34:14 And usually in my company, I think it’s a big advantage, we really collaborate a lot. Compared to other companies, I see there’s a lot more collaboration and it’s and we have quite a variety of profiles. That’s really a strength. It’s really strange it’s not like the more … Normally, there’s just one consultant that works for you and has to take care of everything.

Dr. Fabio Gori 34:34 In our case, it’s like you’re hiring a team directly. It’s very excellent. And sometimes, of course, I visit clients at some kind of business meetings with potential clients or like following up. But I will say that it’s hard to find a standard day. Because sometimes, I also might start to writing grants. For example, applying for grants or tenders.

Dr. Fabio Gori 35:02 What I like about my job is the variety. I can swing between like, I don’t know, having meetings with clients to writing proposals, updating the website, making presentations or like developing the model. It’s hard to pinpoint. With the pandemic, I will say it is quite deep for me. The way I handled the time is very different. Initially, I was working so much at some point on the weekend, I was just dying.

Dr. Fabio Gori 35:36 Now, I found the method that I calculated for each anchor time at work; how much breaks I’m going to take. And then at the end, I think I’m quite efficient, because I removed the distraction. Because as soon as start feeling tired, take a break, and have a little break. And so, maybe what’s going to happen that I work for a longer period, I mean, I may spread my working time over a longer period. But at least, I know the performance better.

Dr. Fabio Gori 36:04 The main advantage for me for the pandemic is the silence. Because I really feel that in the past, I was working in a noisy environment that it was killing me. I could not concentrate at all. And the silence really simplifies my life so much, from this point of view. The issue that I had recently, they were working on something dealing with devices.

Dr. Fabio Gori 36:24 And so, I was doing more of the science part, but I was forced to do more like the software algorithm part. And there was a bit of an issue because I couldn’t ask a colleague to please help me directly. There was a bottleneck. The bottleneck, I will say that this interaction sometimes; it could become difficult. In this case, because it was like a physical constraint, there was a device.

Dr. Fabio Gori 36:47 I couldn’t teleport it to Amsterdam to my colleague. But for the other point of view, like sharing the code, looking at the code together, talking. I think in a few weeks, we made a transition and I think it’s fine.

Natalia Bielczyk 37:01 I think that this pandemic could, in the long-run, change the working patterns for a lot of professionals. I was used to working from home because I used to do that before. But all those who are not, it was like a shock. Like throwing you into the water, if you have to stop commuting to work and start working from home.

Natalia Bielczyk 37:28 But I think after a few months, we all got used to it. And I think many people found unexpected pleasure from working from home. And I think after the pandemic, even when it’s completely answered. I think it will be more common to have that lifestyle to bring up this part of the work home and at least partly work from home.

Natalia Bielczyk 37:50 Many companies already … There is a big crisis right now on the property rental market, because many companies completely resigned from offices. Just because they realize that they don’t need them. Especially in data science and all these disciplines that don’t really require any like wet labs or anything, you know, that is physically there.

Natalia Bielczyk 38:15 I think this situation will change the job market forever probably. Maybe you will be able to work from home for most of the week after the pandemic. Because I think many employers are thinking of reducing their costs by partially or fully moving to working remotely after the pandemic.

Dr. Fabio Gori 38:36 A very recent company is working fully remotely like Elasticsearch. It already existed, but of course, it’s a jump or that in theory could do it. But in practice, they didn’t want to take the risk. Now they were forced to take this risk and make this jump. I have to say after a few months, I also see the downside, so I really miss my colleagues honestly; like on a personal level. Also, because I live by myself; I’m really like isolated. And so, this kind of personal touch, and I really like my colleagues, so that’s something I miss.

Dr. Fabio Gori 38:36 But in my day, setting will be like spending two, three days in the office and the rest home. When you really have one thing to do, and you have to focus on that one, you know that you don’t need that much interaction with somebody; like working on the same device especially. In that case, the possibility of working from home really makes a difference.

Dr. Fabio Gori 39:27 But it is something that I could do, for example, during my post doctorate. There was some work in the building, so it was super noisy; for a while I was working from home. But I had quite some freedom at times, so I could really say, ‘Okay, I’m going to work remotely’, and it was fine. I mean, especially for this concentration, when you need deep focus, it can make the difference if you don’t have kids at all.

Natalia Bielczyk 39:55 Good point. Coming back to the jobs in data science, I would like to ask you about the whole job search process. Because my experience is that, and also many people I know, once you go to LinkedIn and start browsing through these data science offers, they all sound the same. You know, they’re always looking for motivated team player, enthusiastic etc.

Natalia Bielczyk 40:25 And they always offer you the same, like flexibility and friendly team, etc. I mean, it’s almost copy pasted, all these jobs sound the same. Do you have any advice for PhDs interested in these types of jobs, in terms of how to choose the right one?

Dr. Fabio Gori 40:47 It’s hard to say. I mean, it’s not that easy. Because until you get in, you don’t realize the situation. But I will say that, I see quite some difference in the description. If I think about when I was looking, I remember like, … Because I was looking more in the bioinformatic part. I was looking at it for industry, like before I even like electronics thing.

Dr. Fabio Gori 41:11 There’s some people like two years ago, we’re already much into deep learning. They only wanted one person with deep learning. Other ones, they were really like linear models/logistic regressions. As far as I remember, there was quite a variability in the things they were asking. Again, it’s also to see the domain. It really depends on the domain.

Dr. Fabio Gori 41:36 Maybe they want it to look more similar. The one job is like, very IT or purely informatics. The ones I was looking for were somehow applied to biologists, some applied to something really concrete. There was always this component with them that change a lot. My first suggestion, try to understand if they have the data. Because there’s people that don’t know really how artificial intelligence data science works.

Dr. Fabio Gori 42:03 And so they think, like you do magic, but they don’t realize how much you’re depending on the data. First, they need to have the data. Second, they need to have like, already hopefully cleaned data. Because a lot of effort goes into cleaning, especially with Bioinformatics; it’s very tough data,

Dr. Fabio Gori 42:23 You really have to do like an immediately designed pipelines for processing, it can take weeks. Second, clean data and test. Third thing, finding expert people that are already there; knowledge about there, so that you can learn from them. These are the criteria in this order, I would say.

Natalia Bielczyk 42:43 If you think about your job search process, if you think all the way back to your academic career. Is there anything that you would do differently?

Dr. Fabio Gori 42:52 I would’ve start searching for a job sooner. I think like after a few months of my postdoc. When I’m felt like that I want to go on with the academia, I would already start looking for a job, for sure. And then, it’s hard to say what else. Maybe curate a bit more my CV … I also realized that … I mean, I tend to super prepare for the interview.

Dr. Fabio Gori 43:23 But thinking that is the only chance to do something is very bad for your performance. Obviously, there will be a second chance to be more relaxed. That’s the main thing. It’s like, check your CV well, and also don’t be scared about the advertisement you see. They tend to ask things that are clearly impossible.

Dr. Fabio Gori 43:46 In the end, you don’t need to match everything, you just need to be the most matching among the applicants. Be bold. From his point of view, you can be bold so you can try to apply. That would be my suggestions., For example, what I think that made me took a while to transition, I was very selective honestly because I was looking for something combining bioinformatics.

Dr. Fabio Gori 44:12 And there wasn’t that much bioinformatics at the time in Europe; outside of the UK. But it was the thing in my CV that didn’t look nice. And so, probably the reason why.

Natalia Bielczyk 44:30 You raised a very important point here. As you said, ‘You don’t have to be perfect, you have to be less imperfect than others’. Many people get intimidated when they see for instance, how many applicants apply through LinkedIn because you can see these numbers. And they’re like, ‘Oh, man, it’s already 100 people. I have no chance.’

Natalia Bielczyk 44:48 But what they don’t realize is that among these 100 people 50% or 60% was completely random, are people who just showed their resumes everywhere. And among the remaining half, at least half of those resumes and motivational letters were just drafted; not very good quality. When you look at the numbers, if you do it well. You still have a fine chance of getting through, at least the first round, and then have a chance to present yourself. These numbers like don’t really tell the truth about your chances. It’s good to take chances.

Dr. Fabio Gori 45:27 There are two things I would like to say. If you’re probably applying to a big company, you should do something that I don’t do. I’m not and applying for anything now. But take your CV and make it in a format that is like a Word document or something that can be easily be extracted. Because there’s also AI for human resources; it’s a growing field and they’re fancier. Mine is a very fancy mix with a lot of text. It looks very nice.

Dr. Fabio Gori 45:56 But then it’s harder for this kind of algorithm to extract the text. Make something that you know that in this big company the text will be extracted. It’s very basic but it could make a difference. But also remember, if you’re not selected, it doesn’t mean that you’re bad. There are so many reasons why. I can say in this one, because I also was on the other side when I was in biotech; I was interviewing people in both of my jobs.

Dr. Fabio Gori 46:23 I can tell you that this selection can depend on so many things. Maybe you’re very good in what they’re doing, but they’re asking for different set of skills. Don’t be afraid. I mean, don’t feel let down. It’s not an exam, it’s different. It’s more likely they’re looking for a match. And also, the turning power of choosing the match that also matters.

Natalia Bielczyk 46:45 That’s very interesting. I skipped that detail about you, that you also recruited in your previous job. Could you tell us a little bit more about that? Also, in your experience, from experience as a recruiter, as a part of the recruiting team. Can you name some deal breakers, some things that made you reject a candidate?

Dr. Fabio Gori 47:18 This is a tough question. Because it’s also hard to detach this one for the people. But if I think about the same, the first job, what we saw for example … There were some like, was bioinformatics. There were some biologists who just kind of did somehow, started using the computer, so they thought they were for bioinformatician but it’s really not.

Dr. Fabio Gori 47:43 And they said,’ Oh my. If it is a postdoc, I could be a Professor.’ I was already thinking like ‘Wow, how these people can be like me MBL’. I don’t know, before I would have worked in the MBL as well. You can find some people that are overestimating. In some cases, they don’t really understand what they are doing. In other cases, there were people very far from the domain that know nothing about bioinformatics or anything.

Dr. Fabio Gori 48:10 Sometimes I had the feeling that somebody was discouraged because the team was more biologist oriented and it was more computational. It really depends. In this case, it was interdisciplinary one so it could be hard to convince someone to join. The other thing is to know the audience, to see which of the two sides are prevailing and show that you’re complimentary to them, but you’re also convincing on their point of view.

Dr. Fabio Gori 48:40 I will say, in my first interview, this first company was really hard. I had to be really prepared to make a presentation. And one mistake that I was doing that’s important, I was giving very scientific presentation. The best thing is to do a mix about yourself and something scientific.

Dr. Fabio Gori 48:57 Something scientific to show what you’re doing and then they will ask you a question about it. But something about yourself because you’re not just for job, you’re more than that. You have to do something between. And so, they asked me quite tough questions, so it’s good to prepare on what you’re presenting, prepared to be asked questions out of your comfort zone.

Dr. Fabio Gori 49:19 What I saw for example, when I was within the second time in the company that I left. I was recruiting my replacement. In this case, I was the one asking the nasty questions. But I really felt something my boss was falling for a guy that for me was a total fraud. My boss was like a bioinformatician coming from biology. He didn’t realize what this guy was talking about neural networks. It was something so obvious and trivial that everybody could see with minimal knowledge.

Dr. Fabio Gori 49:50 You can also complain your role like if somebody is lacking. If you do a play a disciplinary role and people are mostly focused on certain direction. You can claim that you know more stuff than the other one and they will not find it out. That’s also playing your role. But I will say be prepared on what you’re doing. Ask yourself, if you’re presenting original work. Ask yourself critical questions about what you’ve done. Try to think about some questions you might get.

Dr. Fabio Gori 50:18 And don’t be afraid to say no/yes. I mean, like it depends on the culture. But I know some of my colleagues were frustrated by some candidate because we’re in Germany, and they were not giving direct answers. As I said, it’s more about matching. You don’t have to think about what is the right answer. Just think about what is the answer that reflects myself, somehow.

Dr. Fabio Gori 50:42 Being true. I mean, if you’re not saying the truth and then you’re hired, then you have to match a lie. It can be tough. That’s it in a nutshell. The other thing I could doubt is that, as I said, this is a double match. Something that I can tell will play out very badly. People that you ask questions, and you really feel that they don’t know what they’re doing.

Dr. Fabio Gori 51:10 I remember one candidate, I just asked something. I just sent her a question. She sent me something like we just met and it was very basic. I think she didn’t even listen to me. Ask a question, technical one. But it wasn’t tough, it wase just for understanding better. More like for trying to see if I was understanding what she was saying, and she replied with some nonsense. Listen to what they ask you, let’s say.

Natalia Bielczyk 51:39 Maybe so, it might have been partially due to stress as well. Because sometimes you have this blank wall in front of your eyes, all of a sudden in the interview.

Dr. Fabio Gori 51:49 But in that case, just say sorry, maybe like you can say, ‘Could you repeat your question? Let me think about it.’ You know, but not give an answer that doesn’t make sense. Not replying is better a wrong reply.

Natalia Bielczyk 52:04 That’s also a lesson. That’s a good one. It’s interesting because I found that silence is not the very best thing. Usually, you have to find something.

Dr. Fabio Gori 52:17 You have to feel like it, also, what they say. If you don’t know something, you should say so. I think it depends on the culture. In some culture, they’re very direct to the question, you might say, ‘I don’t know this one’. But then you can come up with some compensation. The more Anglo-Saxon culture, we should come up with a compensation for something that is true.

Dr. Fabio Gori 52:37 In this case, it was like a technical question. It’s not something like, ‘Is my understanding of this thing correct?’ And then you should say like, yes or no. There is no alternative, you know. You can ask somebody to rephrase the question. But coming up with an answer that is neither yes or no, it has nothing to do with the question. It does not play well in your other direction.

Natalia Bielczyk 52:38 You already gave us plenty of information and plenty of great advice. The last thing I would like to ask you. What would you advise to all those people who are thinking about data science, and it sounds very interesting for them, but they didn’t requalify for this field yet? And they are thinking of taking online courses or maybe taking additional studies to start data science jobs.

Natalia Bielczyk 53:34 What would you advise them, which direction is worth it to go today? At this point, when we were talking about this, the amount of data scientists and the supply of data scientists on the market is growing so fast. Is it still a good idea to start?

Dr. Fabio Gori 53:56 As I said, when a friend asked me, she told me she wanted to move. I told her to just become a product manager. Maybe because all the people with a background in biology, or like biomedical things, they end up doing Product Manager. I can say that adding a bit of education in data science is helpful even if you don’t do it as a scientist.

Dr. Fabio Gori 54:16 Because it then allows you to have an understanding of things that are going on. The main problem that I had when I was in this biotech world is that these people, that I was dealing with, had no concept without automation. Bioinformatics, AI comes after the automation and there is a lot of stuff to do in the field of automation.

Dr. Fabio Gori 54:40 And I will say that it’s more important to understand like these IT things; the computational programming works. Because this is valuable, wherever direction you are going to do. And if you want to … Even if you’re going to be a project manager, for example, because I tell you, I really didn’t know and it wasn’t a big limitation.

Dr. Fabio Gori 55:00 I will say that if you really want to move to data science, I’m not discouraging you. I mean, learning more sciences is helpful anyway. As I said, it is something that will combine with your skill. If you really want to … I would say, if you’re like a PhD, try to look for papers of combining your domain expertise where people use data science, AI, machine learning, on things related to your domain expertise. That’s the best thing.

Dr. Fabio Gori 55:26 Because your domain expertise is your main asset. If you can combine it with a science, that is fantastic. Look for that. Also, what I saw is that sometimes the lack of knowledge in methods, could be compensated by knowledge in domain. If you really want to transition, go through something as close as possible to your domain. And even for learning, the good way is to look at papers when they use these methods on your domain.

Natalia Bielczyk 55:56 Thank you so much, Fabio. Thank you so much for all your insights and for being with us today. And for everyone who is watching this episode, I’d like to mention that. If you guys would like to get more of this content, then please subscribe to this channel. And of course, leave us any comments and questions you might have and we will diligently answer all of them. Thank you so much and till next time.