About halfway through my Ph.d. I realized that I probably did not want to stay in academia once I finished. Two things made me realize this. Firstly, I was spending half of the year in Paris while my girlfriend stayed in Denmark. This was not particularly good for our relationship and a career in academia would certainly entail more stays abroad. Secondly, I came to realize that while I liked doing the heavy mathematical analysis of data, I didn’t care much for the subject I studied, which was supra-molecular structures in liquids. I still had 18 months to go through, and I felt like a needed a plan for the post-doc, which didn’t include a post-doc position.
My brother had worked as a software engineer for a couple of years and I had always secretly envied his work environment – free new computers, cool offices, and excellent coffee, and a great salary. My only programming experience was with MatLab, so I did not think I could become a software engineer, but there was a growing hype around data science, machine learning, and artificial intelligence. This seemed like a more promising way in. After looking into the hype, I realized that the work I had been doing as an experimental physicist wasn’t that far removed from the techniques used in data science and machine learning – experimental physics is a data-driven science after all. Since I still had about 18 months left of my Ph.d. position I decided to read up on data science and machine learning. This involved Coursera courses (e.g. this, an earlier version of this and this), a series of small project, and getting experience with Python.
Fast forward 18 months. I handed in my thesis late December 2016. My defense would take place at the beginning of March 2017, so I had three months to find a job. I applied for about eight positions. Got interviewed for about four of them and ended up taking a position as Machine Learning Engineer at a company that develops clinical support systems for heart clinics – a position I started in may this year.
Now I’m about seven months into this position and so far I’m very happy with my decision to transition away from academia. I spent the first three months working on a machine learning model that use data from patients’ cardiac devices to predict severe cardiac events at least one day in advance. Unfortunately, we were unable to reach a satisfactory level of precision and recall with the data accessible to us. Because of this, we are currently focusing on building up the Data Science equivalent of Maslow’s hierarchy of needs to get a better foundation for our machine learning.
This means that I have been doing a lot of database design and writing small programs that extract and transforms our raw data. This has been a lot of fun and we have made enough progress that I have had time to get into Tableau as well. In the beginning of next year, we will embed Tableau dashboards on the webpage used by the clinicians, at which point I will have worked on the full data stack (extraction, transformation, loading into databases, and analytics). Having seen a little everything along the way feels very empowering even though the road itself was very frustrating at times. Hopefully next year we will have enough control over the data flow that it will make sense to start working on machine learning again.
Along the way, I have focused quite a lot on becoming a better programmer. I find that I really enjoy trying to make my code simple and readable. I read Code Complete, Clean Code and watched some of Katrina Owen’s talks on refactoring. Code complete is a very long book, but it was very useful for someone new to software engineering, because it covers a lot of ground and assumes no prior knowledge. Clean code is quite short, but it served as a good refresher of many of the principles covered in Code Complete and the refactoring examples are very useful. Katrina Owens’s talks are fun and it is nice to see refactoring in action. I still have a very long way to go – The Pragmatic Programmer still sits untouched on the shelf – but I’m making good progress.
When I got back from my last stay in Paris in the middle of 2016 my girlfriend and I moved from Odense, where she had been studying medicine, to Greater Copenhagen. We knew that we wanted to buy a home, but we could not afford one – at least not one that we liked. She had just finished her studies and I had used most of my salary on a small but extremely expensive room in Paris and flights between Denmark and France every other weekend. So we rented a small house in the Suburbs of Copenhagen instead. This was cheap enough that we could still save up to buy our own at one point. Ideally, we wanted to own a small house with a garden, and we had pretty strict requirements in terms of location as well. Naturally, this meant that most of the houses that met our requirements tended to be expensive. The houses we would actually afford usually needed a heavy renovation, which we then couldn’t afford. While we were saving up the prices on the housing market kept on increasing, so we could barely keep up. One day I finally had enough, I was feeling stress about my coming home, and I hadn’t even bought one yet. We decided that the whole thing was silly and that the house would have to wait. And that’s how we ended up buying a 60 square meter (645 square foot) apartment when we sat out to buy a house. We will be moving in March next year and so far I’m very happy with our choice. We will be able to keep traveling and generally have a sound budget.
In October we went back to Paris to see one of my French colleagues defend his thesis. I knew that I wanted to propose to my girlfriend and doing it in France seemed like a good choice because of our history with the Country. I ended up proposing at the top of a Medieval tower in the woods of Fontainebleau outside of Paris. Luckily she said yes.
On the personal front, I think we will be enjoying settling into our new apartment. On the professional front, I suspect that 2018 will be a very exciting year. I will be getting a couple of new colleagues and I plan to write a book with my brother on making small machine learning projects – a book I felt I needed when I first got into this stuff. I will continue to work on my programming skill, and would like to get into functional programming. I am looking forward to the new year.