Kirill Eremenko: This is episode number 313 with Data Analyst Marco Caviezel.
Kirill Eremenko: Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, Data Science Coach and Lifestyle Entrepreneur. Each week we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex, simple.
Kirill Eremenko: This episode is brought to you by SuperDataScience, our online membership platform for learning data science at any level. We’ve got over two-and-a-half thousand video tutorials, over 200 hours of content and 30+ courses with new courses being on average once per month. You can get access to all of this today just by becoming a SuperDataScience member. There is no strings attached, you just need to go to SuperDataScience.com and sign up there. Cancel at any time. In addition, with your membership, you get access to any new courses that we release, plus all the bonuses associated with them. And of course there are many additional features that are in place or are being put in place as we speak, such as a Slack channel for members where you can already today, connect with other data scientists all over the world or in your location and discuss different topics such as artificial intelligence, machine learning, data science, visualization and more. Or just hang out in the pizza room and have random chats with fellow data scientists.
Kirill Eremenko: Also another feature of the SuperDataScience platform is Office Hours, where every week we invite valuable guests in the space of data science and interrogate them about their techniques, about their methodologies in the space of data science. And you actually get a presentation from the guest and you get an opportunity to ask Q&A at the end. And in some of our Office Hours, we just present some of the most valuable techniques that our hosts think are going to be valuable to you. So all of that and more you get as part of your membership at SuperDataScience. So don’t hold off, sign up today at www.SuperDataScience.com. Secure your membership and take your data science skills to the next level.
Kirill Eremenko: Welcome back to the SuperDataScience podcast, ladies and gentlemen. I’m super pumped to have you back here on the show. And today’s guest is special to me because we met in person in Switzerland a couple of weeks ago. So sometimes when we travel, Hadelin and I send out emails, you may have seen these come through. Previously, in 2017, we’d send them to everybody, now we have a bit of a better segmentation in the list, so we are able to send them specifically to different regions where we are going to be present. So you may have seen these, where we invite students to dinners. And in 2017 we actually had quite a few catch ups. You can actually keep track of all these dinners and all these travels that we do, which we do share, we invite people to, at DataScienceRoadshow.com, so www.DataScienceRoadshow.com and you can find photos there. So recently I had a catch up in Switzerland and I had another one, a very cool catch up in Rome. So quite some interesting photos you can find there.
Kirill Eremenko: Anyway, back to Switzerland. So I think this was the end of July I was in Switzerland, I sent out the email, I did send it out quite late so only two people were able to make it, Marco and Stephan. And both had amazing, very interesting stories, but Marco’s really touched my heart and fascinated what he was able to accomplish through online education, through online courses and I decided to invite him to the podcast. So in today’s episode, what you will find out, you will find out how Marco uses data science in neuroscience, in his neuroscience research. So he comes from a background of psychology and neuroscience. Then how and why he decided that he wants to move into the industry and become a data scientist in a company and how he went about it. What kind of courses he took, what he got from those courses and you’ll hear a very special story that indeed all the knowledge that he got from online education, through our courses on Udemy.
Kirill Eremenko: He took the machine learning A-Z course, he took the Tableau A-Z course, he took the SQL and Database Design course and the Deep Learning A-Z course. From those course that he took online, plus some other courses of course as well, he got enough knowledge to get the interest from employers and secure interviews and be successful at those interviews. You’ll actually hear a story of how he, under pressure, created some interesting visualizations. And then you will also hear what he’s currently doing and how his work in the industry differs from the work he was doing in research. So you’ll get a very interesting comparison from data science in research and data science in the industry and also how to transition from one to the other, or how to use online education to secure a job in data science. Very inspiring story, can’t wait for you to check out this episode. So without further ado, let’s get started and please welcome data analyst from Switzerland, Marco Caviezel.
Kirill Eremenko: Welcome back to the SuperDataScience podcast, ladies and gentlemen. I’m super excited to have you on the show. Today we’ve got a very special guest, Marco Caviezel, calling in from Basel. Marco, how are you going?
Marco Caviezel: Hi Kirill, I’m doing very great. How are you?
Kirill Eremenko: Fantastic Marco, I’m doing amazing as well. It’s been a while, hasn’t it? Since we caught up?
Marco Caviezel: Yeah, yeah. We first met at a dinner actually, in Zurich. So you were here for your business reason I think and you, yeah, I got this email from SuperDataScience where I was hooked up and then we connected and we had a fantastic dinner, it was great.
Kirill Eremenko: Yeah, we went … thank you for coming down from Basel to Zurich. We went to this restaurant, if anybody’s in Switzerland, particularly in Zurich, it’s called Hiltl, right? Hiltl. It’s a vegetarian buffet, it’s amazing. Every time I go to Zurich, I have to go to that restaurant.
Marco Caviezel: Yeah. It was the first time there, but it was really fantastic. So I will definitely go back if I’m in Zurich.
Kirill Eremenko: So Basel is a bit north of Zurich, right?
Marco Caviezel: Yeah, that’s correct. Yeah, so it’s on the border to France and Germany actually.
Kirill Eremenko: Oh really? I love Switzerland so much for the variety of regions, depending on which region you’re in. If you’re in Zurich, which is kind of in east Switzerland, if you’re in Zurich and you’re closer to the German border, people mostly speak German. As soon as you get on a train and you go for two hours, this happened to me, I got on a train I think from Zurich to … yeah, from Zurich to Geneva, Geneva is in the west, so you’re on the train for two hours or so and when you get on everybody’s talking German, when you get off everybody’s talking French. It’s like, what? And some of those people are the same people that got on, but they just switch languages when they get into the French part.
Marco Caviezel: Yeah, so we actually have four official languages in Switzerland, which is French, German, Italian and Rhaeto-Romansh, which is an ancient language. Not many people speak it anymore. So yeah. But these are-
Kirill Eremenko: But you have a few areas where people just speak Romansh as well, right?
Marco Caviezel: Yeah, true. Yeah, but usually they can also speak other languages, so other official languages like Italian or French or German.
Kirill Eremenko: What do you speak?
Marco Caviezel: So we learned French in school, I had French for about eight years. And English and my mother tongue is German. So basically it’s Swiss-German, it’s kind of a dialect of the normal German, let’s say it like this.
Kirill Eremenko: It’s interesting because what I learned when I was in Zurich, in that area, Swiss-German doesn’t actually have a written form. When you write you write in normal German, or it’s called High German-
Marco Caviezel: Yeah, exactly.
Kirill Eremenko: … but when you talk you talk in Swiss-German, which is completely different. Well not completely different, but very hard for a German person to understand Swiss-German, is that right?
Marco Caviezel: Yeah, that’s true. So that’s kind of … yeah, an advantage for the Swiss people because we understand maybe all German speaking or people from Germany perfectly, because we speak High German in school. And also what’s different, we have kind of different dialects from … so we have cantons, so Switzerland has a lot of cantons, like areas or regions and they also have different languages, or different dialects, but we understand each other. But you can really differentiate each other, so from where you’re coming originally.
Kirill Eremenko: And that’s really cool, because when … I don’t know any other language that has a spoken form, but doesn’t have a written form. I’m sure there are out there, but this is a language in one of the central countries in Europe, a very influential country. Imagine if I want to translate something into Swiss-German, I go on Google Translate and I can’t even type it up because it doesn’t have a written form. So how do you translate something into Swiss-German or out of Swiss-German? It completely destroys this whole concept of I can just use Google Translate for absolutely anything.
Marco Caviezel: Yeah, yeah. I totally agree. So this is also one of the … yeah, so for example, Siri just has no idea of Swiss-German. So you can’t speak to natural language processing in Swiss-German, so you have to use High German or you switch to English, so that’s kind of fun.
Kirill Eremenko: Wow. So when artificial intelligence takes over the world, it’ll have real trouble getting into Switzerland.
Marco Caviezel: Yeah, yeah, definitely. It wouldn’t understand us.
Kirill Eremenko: Your country is famous for that. Like before the whole bunker … every house, every building in Switzerland, as far as I know and I’ve seen and I’ve been told, has a bunker under it. Is that true?
Marco Caviezel: Many houses have. There are some, I think … so our house doesn’t have one, so don’t attack us please, but we also have a plan where to go in case of emergency. So there is a bed for every citizen in Switzerland in a bunker actually, yeah.
Kirill Eremenko: Wow, wow. Okay. Very well prepared for then for the third world war or whatever.
Marco Caviezel: Yeah.
Kirill Eremenko: Okay. Well Marco, I have to … to kick this podcast off, I have to congratulate you because you did something that a lot of people will see as impossible. You moved from a completely unrelated field to data science, into data science, and you actually got your first job as … you told me just before the podcast you’ve done two weeks there. So huge congratulations to you, that’s a massive, massive accomplishment.
Marco Caviezel: Yeah it’s really … thank you very much. It’s fantastic to be now officially working with data all the time. It’s fantastic, yeah. And it’s a huge step for me as well. Yeah. Also I think thanks to also you, because of your courses, I did a lot of your courses online on Udemy and it really helped me to get into data science. Especially machine learning and deep learning and it also helped in the interviews to show the degrees of Udemy or the certificate of completion. Yeah, that was really a great help.
Kirill Eremenko: Fantastic. Well thank you for that, I really appreciate the positive feedback and the comments. I wanted to actually stop on this a little bit more, not for the sake of how specifically I helped you or the courses we’ve created helped you, but more to understand and for listeners out there who are doubting themselves, who might be thinking I’ve never done data science, I studied something completely different, an arts degree, I’m curious about the space but there’s no way that I can possibly get a job through just online education. Tell us a bit more about that. So what field did you come from into data science and how was this journey of deciding I want to do data science?
Marco Caviezel: Yeah. So as you said I started, I did my bachelor’s degree in psychology in Basel. I was focusing there on economic psychology, this was the meta-topic I was looking into. And then I got into the topic of neuroscience, actually my masters degree, which I also did here in Basel. So then I came in touch with first programming languages and more computer and technical things, let’s put it like that. So, for example, I learned a little bit how to use MATLAB to design experiments and stuff.
Kirill Eremenko: For neuroscience?
Marco Caviezel: For neuroscience, exactly, yeah.
Kirill Eremenko: I would … no negative connotation for neuroscience, but I would expect that’s a much lighter application of technology than the data science that you’re doing now. Would that be a correct statement?
Marco Caviezel: Well I got into really, some neuroimaging in my PhD, so after my masters I started my PhD in neuroscience also. If you go into the neuroimaging, I don’t know if you’re familiar with MRI, for example, so magnetic resonance tomography. So what we are doing, or what people are doing in this field is basically they are scanning the brain while it’s doing something and they try to interpret this. And you can think that you have really a lot of data if you do this. So from this perspective, it is really data driven, or it should be data driven in my opinion. So the more I got into the field of neuroimaging, the more I dealt with data as well in research.
Kirill Eremenko: Okay, so is that how your passion for data developed?
Marco Caviezel: Yeah, yeah, definitely. So if you look at the neuroimaging field, the community is really great. So it’s a really huge community and they’re open-minded, most of them are going for open science, so you can really look at code of other people on GitHub and try to implement it on your project. So everybody feels really welcomed in this field and this is fantastic, to get into this kind of data science field. And they are really doing crazy things with deep learning and trying to understand how the brain works, which is a really complex matter.
Kirill Eremenko: Okay. And can you share an example of an experiment you’ve done? You’re doing your PhD now, right? Is that right?
Marco Caviezel: Yeah, yeah.
Kirill Eremenko: Is there any examples of experiments that you can share with us where you used data in neuroscience for, I don’t know, in these MRIs or imaging or any other type of approach? I’d be really curious to know how data science is used in research.
Marco Caviezel: So for example, in one project we were doing functional MRI, so you can imagine that we are basically taking a 3D video of the brain while you are doing stuff on a computer. And we can measure the blood flow indirectly and see what brain areas are active during a certain task. So this is what one project was like, on a broad …
Kirill Eremenko: What was the experiment there?
Marco Caviezel: So we were showing people images of objects on a screen and they had to compare them to a cue they saw before. And either it was completely the same, it was completely a different object or it was just a tiny little bit different than the original cue. And we tried to show what brain areas are active in this, only in this, when the cue or the stimulus is only very, very lightly different from the cue. And this is a process which happens in the hippocampus, which is also used in learning. So I was in the center of old-age psychiatry and we were trying to find a pattern, or an experimental approach to activate this hippocampus. Because in Alzheimer’s disease, for example, this part of the brain just is gone after a while and then you can’t learn anymore and that was the reason, for example, why we are investigating this part of the brain.
Kirill Eremenko: Wow, very interesting. And did you have any significant results from this?
Marco Caviezel: Yeah, definitely. So we are actually now publishing, or trying to publish one study about this. So in the first study we are only using or measuring young, healthy subjects to look if this paradigm which we developed works at all. And in the second step, we do this with old, healthy subjects to see the normal aging process and then we compare this aging process to Alzheimer’s patients and see what is different, basically.
Kirill Eremenko: Okay, got you. What kind of techniques from data science do you use for the analysis?
Marco Caviezel: So this is a general linear model, so what you first do is … so every subject is doing this experiment and then you are doing a first level analysis. So you’re looking for this individual subject, which brain areas are active. And then on a second level, you use this … so there’s contrast images, what they are called, so the result of the first level are called contrast images and you take these and compare them in a group. But you have to do really a bunch of pre-processing steps to get to this second level and really have a definitive result that you can present.
Kirill Eremenko: Okay. Okay. And all of that is in the scope of a general linear model.
Marco Caviezel: Yes. So you have, for every subject you have a general linear model, which is basically just a regression analysis, right? So just a linear regression and your dependent variable are your images or the images of the MRI scanner, which are basically also just ones and zeros. And then you have your model with your variables you’re defining, so you have to somehow get the images and the behavioral data together. So the dependent variable would be the images and we try to explain by your behavioral data these images. And then you do a contrast analysis, which means you are looking at the betas of your independent variables and you then see what areas are explained by these specific independent variables. It’s kind of complex if you have to describe it without writing it.
Kirill Eremenko: Okay. But tell me this, how do you put an image into a general linear model?
Marco Caviezel: So it’s basically a 4D matrix, so you have … it’s called voxels, so it’s like a pixel, but in 3D. So if you imagine just a normal image and you have a pixel, this pixel is just … and it’s a black and white image, let’s say. This would be one or zero, right? And if you put it all together, then you get your image in total. But the thing is, in fMRI, you don’t have 2D, which would be a normal image. So you basically have a 4D actually. So you have the time because you’re taking kind of a video and you also do this in 3D. So you have this 4D matrix and then for every voxel, you’re doing the general linear model.
Kirill Eremenko: Oh, for every voxel, okay.
Marco Caviezel: Yeah, yeah. For every-
Kirill Eremenko: Every point in your 4D matrix.
Marco Caviezel: Yeah, exactly, for every point. Yeah, yeah. And then you have the problem, of course, of multiple testings and then you have to correct for that. And yeah, so …
Kirill Eremenko: Okay, okay. That part I understand. Then can you explain, how do you put the behavior into a GLM? You said your independent variables are the behaviors of the person you test.
Marco Caviezel: Yeah, exactly.
Kirill Eremenko: And the dependent is the image that we have as a result. Great. I understand how you put the voxels individually, you have a separate linear model for every voxel.
Marco Caviezel: Yeah, exactly.
Kirill Eremenko: But how do you put the … or you put every single voxel separately into the linear model, but how do you put the behaviors in there?
Marco Caviezel: So what you do is … so you have your onset times of the events, so you have to be really precise to the millisecond down. And then you know the onset times of, let’s say, an image which, or a stimulus which is just a little, a tiny little bit different than the cue.
Kirill Eremenko: The original cue that you provide.
Marco Caviezel: The original cue, yeah. And then what you have, basically you have for every onset, for every stimulus you are presenting, you have your onset and then you have … what you are doing then is you convulse these … so basically the brain is not working on and off, right? So it’s a hemodynamic response we are modeling, so how the blood flow is affecting this area of the brain. So we have this hemodynamic response and we have … so basically we have the Stick function and we use this hemodynamic response to try to say how the brain image should behave. And this is how we try to model-
Kirill Eremenko: I thought behavioral data was something the person does, how they maybe click a button, or how they react, those things?
Marco Caviezel: You can choose whatever. Actually you can choose whatever you’d like as an onset time. You can choose the onset on the click of a button or also the onset of a stimulus, for example. So this millisecond where you see the image. And then in the brain, something happens. But I totally agree, it depends what you take as an onset time.
Kirill Eremenko: Okay, well let’s say we take as the onset time when the person clicks a button, right? So that’s a behavior. How do you put that behavior into the model?
Marco Caviezel: So you know exactly when this person pressed this button or clicked the button, right? And then you try to model this hemodynamic response. So you have your behavioral data and let’s say you have your clicks and every stimulus time is one independent variable of your model. So let’s put the independent variable of the model is if you click on a stimulus which is the same as the cue, the independent variable number two is the stimulus when you click on a completely different image and independent variable number three would be if you click on a stimulus which is just a tiny little bit different to the cue. And then you try to model it with these three parameters, every voxel, basically and then you are looking, using contrast analysis. What part of the variance is explained by independent variable number one, which would be the same condition. And if you look at independent variable number two, you would see the completely different and number three the tiny little bit off. And then you can see where in the whole brain, or which voxel activates the most based on these independent variables.
Kirill Eremenko: Okay, okay. Fantastic. I understand now. Thank you. So I got you. On the left you have what button they clicked, on the right you have the voxel of … so every single voxel of that image. As I understand, the model is quite simple, but computationally it’ll be quite heavy because you have to do this for every single voxel in your scan.
Marco Caviezel: Exactly. Yeah.
Kirill Eremenko: Yeah, okay.
Marco Caviezel: That’s really … and you have this for every subject. And then you go to the second level and then where you put all the results or all … so for each subject you then get … for each subject and for each condition you get one beta image with the brain areas which are activated for this condition and then you take all of these images of many people or several persons and put them together and look at what is on average, or what is the group effect. So this is how you combine then in the second level.
Kirill Eremenko: Okay. And one other thing that I’m curious about, you yourself know if that image that you show them is the same as the cue, different to the cue or slightly different to the cue.
Marco Caviezel: Yeah, yeah, exactly. I know that.
Kirill Eremenko: So they might make a mistake, right? They might click the wrong button and so you’re putting that into the model, what they clicked and what the fMRI showed. My question is, does the actual, the truth, does that go into the model as well? Do you somehow factor in that knowledge about what actually was the case? Was it the same image, different or slightly different? Does that go in as a parameter in the model at all?
Marco Caviezel: Yeah. You can use this, the performance, as a … you can look at what brain areas are correlated with the performance, for example, this is what you can do. And then you just look at this independent variable. Or if you just put it in the model and you treat it as a nuisance parameter, then you would correct for it. Or you just … the other option would be that you just don’t take these onsets which were not correct into the model. Just leave them out.
Kirill Eremenko: Got you.
Marco Caviezel: There are different ways how to do it and, yeah, this is one problem. There are so many ways that it is hard to reproduce studies from other groups. Yeah.
Kirill Eremenko: Okay, I understand. So it sounds like very interesting work. At what point did you decide that you don’t want to do research, you want to move into the industry and become a data scientist in the industry?
Marco Caviezel: So it was in research, I would say … I don’t know if it’s everywhere like this, but I guess in Switzerland you don’t have this secure job, or you don’t know what … so you basically often have jobs, or a contract for three years max and this is one point. And the second one is that I didn’t want to move away from Switzerland. And in research, if you really want to do this your whole life, it’s basically mandatory to travel and do your postdoc in different labs and different countries. So yeah, I have a family now and it’s just not-
Kirill Eremenko: By the way congrats, on your daughter.
Marco Caviezel: Thanks. Yeah, yeah, it’s fantastic. Yeah. It’s also different. Everything now changes.
Kirill Eremenko: Yeah. Okay. And so you decided that you want to be stable in Switzerland.
Marco Caviezel: Exactly.
Kirill Eremenko: And data science was an interesting way for you to develop your career. Why data science?
Marco Caviezel: So I took some courses on machine learning, so this is how it started actually. So we were writing this paper and after writing this paper I needed a little break and I thought I wanted to do something else, or a little bit else, which is also related to data. And there’s also machine learning in neuroscience. There are lots of research groups using this. And so I came, I went to Udemy and saw this course which kind of changed my life.
Kirill Eremenko: Which course was that?
Marco Caviezel: So it was the Machine Learning A-Z course.
Kirill Eremenko: The one we, the big one that we created?
Marco Caviezel: Yeah, yeah, yeah. So I thought, yeah, it’s 12 euros or so, it’s not as much, I’d just give it a try and I started it and I thought it’s really … yeah, I learned so much in this course. I also had some courses at university, but they are not that well structured to be honest.
Kirill Eremenko: Oh awesome. That’s great to hear.
Marco Caviezel: And also I really enjoyed that you can code with Hadelin at that moment, who is also giving the course, I think, together with you. And I really made progress and understand what you’re doing and get all these great intuitions. So this is why, or how it all started, basically, to get into data science. Although I think also, well data science is, for me, it’s not a really well defined area, I think. So that’s why … if somebody’s asking me why you’re completely switching from psychology to data analyst or data science, then I would say yeah, it is kind of a big change but the methods itself, if you understand them and you know how to interpret results and try to visualize things, this is also in psychology, or especially in neuroimaging. It’s a big part and you can transfer this methodology into every field also in the industry, as I experience right now.
Kirill Eremenko: But what does this move, at the same time moving from research into industry, how are they different? For somebody who doesn’t know one or the other, how would you compare research and industry?
Marco Caviezel: So research, I think it depends on what research you are doing. But I think in research you are more interested in … so the things you are studying don’t have to be very practical, right? So the usability can be questionable. For example, the experiments that I told you before, our design or method we used. So I don’t know if this really helps humanity a lot, or someone really a lot. So it’s basic research, of course, but the application is sometimes questionable. On the other hand, in industry you really are focused to deploy your models or deploy everything you are doing because at the end you want to save some money or make money, right? Or save time or make work easier for people. And so they really have a benefit of this. And I think this is one of … for me that’s why I like that, it’s one of the big differences.
Kirill Eremenko: Okay, got you. So can you tell me, please, how … was it scary or was it hard to apply for jobs? What did you do once you decided I want to go work in the industry, I want to be a data scientist? What did you do from there? Did you take more courses, did you just start applying, did you, I don’t know, ask around your friends and colleagues and so on?
Marco Caviezel: So I just started applying. Yeah. I got some … so you have to deal with rejections, but you don’t … yeah, just shake them off and go on. So it is … I guess it is really hard to get into it if you’re not in the field. Yeah, so if you are not in the industry I think it’s really important that you try to take the lead of the conversation and really focus on what you can do and not on what you cannot do. So I think if you take the conversation and just talk them through all your successes you had in your career, even if it’s smaller things but related to data, even if you’re good in Excel or in Pivot table or in, I don’t know, a sports club and you’re doing the database with all the members, just tell what experience you have. So I think this helped a lot.
Kirill Eremenko: Okay. That’s a very fair point, you’ve got to demonstrate that experience. But at the same time I’m looking at your LinkedIn and I can see that you not only did the machine learning A-Z course, which you completed. Congratulations, it’s a 40 hour course.
Marco Caviezel: Yeah, yeah. But it’s worth it. I can recommend it to everybody.
Kirill Eremenko: That’s awesome. Then you also did the Tableau 10 course, you did the Deep Learning A-Z course, you did the SQL and Database design A-Z course, so it looks like you were trying out different areas of data science. Tell us a bit about that, were you not confident that it’s going to be machine learning, or was that necessary in terms of the jobs you were looking for? Why did you go from machine learning to visualization to deep learning to SQL and database design?
Marco Caviezel: So what I … if you look at, so what I did is I looked at what companies needed and so, for example, SQL is just a thing almost every … at least in Switzerland, in every application it says SQL is kind of mandatory. So I just went there and did a course on SQL. I would say machine learning and deep learning is really my passion, so I was really, it was fun to do them. And also Tableau, I think it’s good to get … so I still use Tableau for really quick analysis, because it’s just drag and drop more or less and you don’t have to code your plots you want to show to people. It also was very handy in research, I have to say.
Kirill Eremenko: Oh okay, how was Tableau handy in research?
Marco Caviezel: Yeah, so for exploratory data analysis, I would say it’s fantastic. They also have the student license where you can get the complete Tableau 10 for free. And that’s really … I think this is great to really quickly look at your results and see, look at trends and see where you want to dig in even more. So this helped a lot.
Kirill Eremenko: Okay. Okay. Got you. So now we’ve seen … so hold on, you took all these courses and then you were applying for jobs. At what point did it come through? Somebody or some company responded, how did that happen? Did you have a selection of offers to go through or was it one company that you really liked that responded to your applications? Tell us a bit about that.
Marco Caviezel: So there was one company, it was a bank in Switzerland, there is lots of banks.
Kirill Eremenko: Like everywhere. You can’t throw a rock in Switzerland and not hit a bank.
Marco Caviezel: That’s true. But at the end I was … I didn’t get the job at the end, although it would be great I believe. But the point there was that they had somebody who was from the field of banking, so he had this industry knowledge which I didn’t have, obviously, coming from neuroscience, it’s something completely different. And then I guess it was kind of the next job I applied to, which is where I am now. So I had two interviews there. On the first one it was with my boss which I have right now and even more senior people, some of them, and this went well. And on the second one I had this little task in Tableau where they gave me half an hour and presented me with some data and just said, “Do some insights,” or, “Give me some ideas, what can you do with this data?” And then I had to present this data and prepare it in half an hour, which was kind of a little bit stressful, but still a good experience.
Kirill Eremenko: So was the knowledge you got from the Tableau 10 A-Z course on Udemy, was that enough to do that challenge?
Marco Caviezel: Yeah, yeah, definitely. Yeah.
Kirill Eremenko: That’s so cool. So that course is only seven hours long. So in seven hours you’ve got enough knowledge and practical experience in order to successfully perform something at an interview in a stressful situation. That’s very, very cool.
Marco Caviezel: Yeah, it was fantastic. Yeah, it’s just you have to sit, or be calm and try to be calm and just think what you can do. And I think Tableau really helps because it’s quite intuitive in everything. And you just try and if it’s not working, you try the next thing. And at the end you have some results which you can present.
Kirill Eremenko: Fantastic. What did they say?
Marco Caviezel: I think they were quite impressed. Because I also used this geography tool where you can plot the maps in Tableau and I think they really liked that one.
Kirill Eremenko: So the investment of $12 into the course paid off?
Marco Caviezel: Yeah, totally. Totally paid off, yeah.
Kirill Eremenko: Oh that’s so cool, awesome. That’s very cool. Okay. And so now you are in this company. I can’t even … I’m looking at it on LinkedIn, I can’t even read the name. Feldschlösschen. Can you pronounce it for me, please? It’s so complicated.
Marco Caviezel: Yeah. It’s Feldschlösschen. So it means field castle.
Kirill Eremenko: Field castle?
Marco Caviezel: And it’s a brewery. It belongs to the Carlsberg group.
Kirill Eremenko: So what do you do there with data science, at a brewery?
Marco Caviezel: So I am basically in the supply chain, or in the logistics. So we have production of the whole beer, we also are the largest beverage supplier in whole of Switzerland. So not only beer, but also all other kinds of non alcoholic drinks. And the next thing, so we do all the supply chain basically by ourself, so it means we produce products or beer in this case and we store them, we distribute them, so it’s … yeah, it’s quite a lot of work to get that bottle of beer in your hand. I didn’t believe how much effort it takes to have a beer just at a restaurant somewhere and to see all the supply chain which needs to happen.
Kirill Eremenko: Okay. And so what kind of data science do you do there in this supply chain?
Marco Caviezel: So we are working with key performance indicators and reporting, so this is one part of the job. So we are visualizing stuff for people that are using them right now and also to senior people and we do this in Power BI, not Tableau anymore. But I think it’s quite transferrable. So if you know Tableau, you’ll really get into Power BI really quick.
Kirill Eremenko: Yeah, for sure.
Marco Caviezel: So we’re doing these reportings, so I also learned now in this to handle SAP, which is … and Microsoft Access, which we are using still. We are thinking actually to migrate to the Azure cloud next year, so this is one big project we are approaching. So far there was no, let’s say, predictive analytics done and this is also one of the reasons why they hired me, I guess, because I have some understanding of time series analysis and some knowledge in Python and R, which they really like. And yeah, that’s one part of what we are doing. And I am also now starting a project with robotic process automatization.
Kirill Eremenko: RPA, yeah?
Marco Caviezel: RPA, exactly. Because there are still some roles or some people doing lots of stuff manually, which we could free up their time with repetitive tasks.
Kirill Eremenko: That’s very different though, to machine learning A-Z and Tableau, it’s another field. It looks like you’re doing … they’re asking you to do a lot of things in different spaces in data science there.
Marco Caviezel: Yeah, I have a kind of freedom to do what I like, or what I can do. I also got into this RPA. So I really like to look at processes you can optimize and yeah, look at how you can do things easier or more efficient. And I think this also comes back from economic psychology, looking at processes and evaluating them. So this is still in the back of my head, I would say. And I think it’s great to be a link between the data, which I really love, but also the people and to see how you can help them, or how you can make their life easier. Yeah, for me, to be this link, it’s really great. And also to know … because I’m genuinely interested in technology and all you can do, so for me data science is not a nine to five job, let’s put it like that.
Marco Caviezel: So I love to read articles and books about it all the time, listen to podcasts and yeah, I think it’s great if you have the ability to know what is possible and also to have kind of a rough estimate of how much time it would need to implement this. So for example, RPA is kind of the thing which is kind of quickly implemented with these tools, while, for example, to automate the complete warehouse would be a huge project, but it would still be possible.
Kirill Eremenko: So you don’t find RPA complex to learn?
Marco Caviezel: So basically I started a week ago to be honest and I’m really doing progress with UiPath, which is also a kind of, it’s really a drag and drop thing, so you can kind of really quickly start to make progress and to make, to save a lot of time and then money in the end, right?
Kirill Eremenko: Okay. Got you. Yeah.
Marco Caviezel: Yeah, yeah. Just we also work with, I don’t know if you know Microsoft Flow, which is in the Office package?
Kirill Eremenko: No.
Marco Caviezel: It’s also kind of an RPA, but in the Microsoft environment. And it’s more or less just for … so as far as I know, it’s more for Office or for the Microsoft environment. You cannot use or go into SAP with Flows. But it’s still great to start and we also did some Flows and it saves a lot of time as well.
Kirill Eremenko: Okay, all right. Cool, got you. Thank you for the description. So we’ve started coming to the end of this recording, or this podcast, I wanted to ask you, what would your … I don’t know, suggestions, ideas, wishes, recommendations, encouragement even, be to those who are listening and maybe they’re unsure if they can get a job in data science? Maybe they feel, for whatever reason, that they’re not good enough, that a job in data science is too hard for them to get. And then here we have an example of you, where you just decided you wanted a job in data science and you started taking online courses. And those certificates you got from Udemy and the knowledge that you got from Udemy was enough to get you through all those interviews and land you this job where you’ve been for two weeks. So what would your words of encouragement and inspiration be to people who are in the same boat as you were in a few months ago?
Marco Caviezel: Yeah. So you just always have to think what is the worst which could happen? You stay in your job, right? So just give it a try, try to apply and with every interview you learn more things. And I think it’s really important not to be afraid of a complex field. So if you just spend enough time and really want something, to learn something, you can definitely learn it. So I never, before the course I took two years ago, I didn’t even know that Python existed and now I think I can more or less use Python for machine learning.
Marco Caviezel: And just don’t be afraid of new things, just go into it and just try to be really genuinely interested in data science. If you have this, I think you have a good chance to go and get a job in this field, because these people, there is a lot of need of such people who are just interested and want to learn things and try out things and are not afraid of failure and just do it again and start all over it. Yeah, I think this is what I would suggest to people who want to get into data science. If you really want it, you can do it, I would say.
Kirill Eremenko: Fantastic. If you really want it, you can do it. Well, thank you very much, Marco, that was great, great words of encouragement. And thanks for coming on the podcast.
Marco Caviezel: Yeah, thank you.
Kirill Eremenko: Really cool story. No problem. Before I let you go though, could you tell us please where is the best way to connect with you if somebody wants to ask some follow up questions or exchange some knowledge or experience in data science or any of the topics you talked about, what’s the best place to find you?
Marco Caviezel: So I am on LinkedIn, I think this is the easiest way to connect. I think you can add it to the show notes?
Kirill Eremenko: Yes, of course.
Marco Caviezel: And just feel free to contact me and add me.
Kirill Eremenko: Have you ever done that yourself? Have you listened to the podcast and connected with someone?
Marco Caviezel: Yeah, of course. Basically I connect to almost all the speakers you’re having on your podcast.
Kirill Eremenko: Fantastic.
Marco Caviezel: And also I had some discussions with people from your podcast. Yeah, it’s really fantastic to grow or to have this network and grow this network. And it really helps also in the news feeds of LinkedIn, or I don’t know how they call it, to see what are the state of yards, things you can do in data science. So yeah, I recommend everybody who has not LinkedIn yet to go there and make it, make an account.
Kirill Eremenko: Fantastic. Yeah, that’s really cool. Have you listened to many podcasts on this show?
Marco Caviezel: Yeah. So basically I kind of stopped to listen to music. Not all the time, but I started … so I decided for myself that if I want to listen to something on a bike or in a train, I want to listen to something useful if I have the energy. And I think the podcasts are a really great way to do this and also I love music, but I do it really if I want to … I don’t just plug in music and listen to music, I do it, I try to focus on it and enjoy the music and not just the background.
Kirill Eremenko: Got you, got you.
Marco Caviezel: Yeah.
Kirill Eremenko: Awesome. Well fantastic. And now you’re on this show and hopefully people got inspired and will connect and ask questions. That’s amazing. I love this feeling of community that happens in these situations.
Marco Caviezel: Yeah, that’s really amazing, that’s fantastic.
Kirill Eremenko: That’s awesome. Okay. And one final question for you today, what’s a book that you can recommend to our listeners that will help them in their careers or in their lives?
Marco Caviezel: What I would say helped me a lot is The Everything Store about Jeff Bezos and Amazon. So it’s kind of an autobiography, but you really have, I really loved his mindset of how you should focus on things that matter to you. And also about the user experience, I think this is really important for everyone, even if you don’t have someone, let’s say, buying a product from you, but you always try to make the best of your work for those people who are working with your results. And this is just my general interpretation of this book. And I really loved to read it, so basically I listened to it as an audio book, but still, I really recommend this one.
Kirill Eremenko: Fantastic. So The Everything Store by Jeff Bezos. All right, Marco, thank so much for coming on the show today and sharing your story with us. I look forward to catching up in Europe hopefully sometime next year.
Marco Caviezel: Yeah, that would be great, yeah. Really look forward to that.
Kirill Eremenko: All right, take care my friend.
Marco Caviezel: Cool.
Kirill Eremenko: Bye.
Marco Caviezel: Thank you, bye.
Kirill Eremenko: So there you have it ladies and gentlemen, that was Marco. I hope you enjoyed this podcast and I hope you are inspired now. How cool was that? My favorite part of this whole episode and this whole story is that Marco is a testament, Marco’s story is a testament to the fact that it is possible. You don’t have to be a genius, you don’t have to have a huge background in computer science and mathematics and statistics. All you have to have is passion and the dedication to learn and you can even find those courses online, you don’t have to go to university. You can, you can definitely get a batchelor or masters in data science, those are popping up around the world. And of course, depending on the person, somebody might want to look into that, but Marco’s story is a testament that, wherever you get your knowledge, even if it’s through online courses, you can get the right knowledge and qualifications and experience.
Kirill Eremenko: What I like about online courses and the way that I choose to create them is that we add practical exercises. We make sure that there’s practical exercises, because those are very important. Not only just to get some theory or understanding of the tool, but actually being thrown into real life, or lifelike situations with data science where you have dirty data sets, you have to clean them, you have to look for different irregularities in the data or connect data sets and then derive the results from there. But you can actually get all of that experience through online education. And then all you have to do is make yourself visible and look for those opportunities, or get those opportunities to look for you.
Kirill Eremenko: As usual with the SuperDataScience podcast, you can get the show notes for this episode at SuperDataScience.com/313. That’s SuperDataScience.com/313. There you’ll get the transcript for this episode, plus any materials that w’vee mentioned in our conversation with Marco. Plus, of course, the URL for Marco’s LinkedIn and URLs where you can contact him and get in touch and ask him any questions that you might have about how you can best structure your career and maybe get that additional inspiration that’s going to push you forward.
Kirill Eremenko: On that note, thank you for being here today. I wish you to build an inspiring story of your own and break into data science routes, sky rocket your career to the next level and I look forward to seeing you back here next time. Until then, happy analyzing.
