I’m still at the company I joined last year. The assignment I was originally hired to work on (predicting severe cardiac events) failed. That meant that I had to figure out what else we could do with our data. So naturally, this year has included a lot of data mugging and exploration.
By the end of last year I had just finished setting up our Data Warehouse – a process that included a lot of data cleaning. So I’m sure that past Mikkel would be surprised to learn that current Mikkel still has not put machine learning into production. This is despite the fact I gave a presentation on building and shipping machine learning products at a conference in France this year.
There are several reasons why this has not happened yet. The first reason is that we simply did not have a particularly good use case – the models I had been building did not perform well enough. Secondly, once we got something that actually worked, the act of getting it into production also proved quite difficult. This is hardly surprising, everyone says that building the models is the easy part. But we are getting close and I think we have the correct people in place to make it happen.
What I ended up focusing on is a tool for exploring electrocardiograms, which I have described here. I really like working on this product because it brings a lot of my interests together:
- The data is similar to some of the stuff I worked on during my Ph.d. It’s a measurement of voltage over time.
- The analysis requires an interesting mixture of classical signal processing and machine learning.
- The final product is nice interactive visualization – one of the things I liked the most about my Ph.d. work was creating pretty figures.
- Finally, when the tool enters production I will have build a machine learning system where I have developed every stage of the pipeline – from raw data to the final interactive visualization. I’m very proud of this, it is something I have been wanting to do ever since I switched away from academia.
We closed a round of funding which means that I have been in charge of hiring a data science colleague. The hiring process was very time consuming: Writing the job application, shifting through the incoming application, and interviewing all takes time. This was a full-time job for about three weeks, but I’m glad I put a lot of effort into it. I believe we got the right candidate, and the process has been very educational.
I’m very excited about getting a colleague I will be working with closely. So far I have been working with students and I really enjoy teaching this stuff and having someone to discuss problems with. However, there is only so much that can be done in 3-4 months by students and interns. Getting an experienced colleague who will have the time to dive deeply into some of the challenges we face will benefit the company greatly.
I got married! Planning a wedding is exhausting, but I am glad that we did it and I enjoy calling Lise my wife.
We took over our new apartment March this year after an intolerable six months of waiting. The new apartment is great and we really enjoy being closer to the capital. The apartment has required a fair deal of renovation and maintenance and I’m very happy with how far we have come this year. There is still a lot of things that would be nice fix up, but it is in a great place now.
One of my goals for this year was to travel abroad three time with my wife. We went to France, Israel and Norway. The plan succeeded despite the fact that we opted to explore Denmark during our summer holidays, which also gave us some more time to work on the apartment. I hope to go abroad three times next year as well.
I feel like my programming skills have improved significantly over the last six months. I’m very happy about this and hope to improve further next year. I have been playing around a lot with D3.js in my spare time, and I’m really enjoying it.
On a seemingly boring note I have revived my retirement saving plan. While the act of setting aside money for retirement is boring and slightly annoying, the reflections the act inspires are interesting. You have to think about the life you want to live in 35-40 years time. My parents and in-laws have been suffering for various health conditions in 2018, which makes one realize how quickly ones ability and mobility can decrease. So while it’s prudent to save for a long life, its also important to do the things you want to do while you are young an healthy. The retirement savings plan should account for this. Picking a savings plan brings you in an introspective mood.
I have already made some comments on my hopes and expectations for next year, but lets finish with some more concrete goals. Generally, I have a feeling that next year will have fewer major events than the previous two years, since this period has included: Finishing my Ph.d., transitioning to industry, buying an apartment, and getting married. My goals for next year will be more low-key, I hope to:
- Get the electrocardiogram analyzer into production. As I mentioned above this would be career milestone for me.
- Go to more Meet Ups. I found a weekly data visualization Meet Up at my old university. I plan to go there 2 time a month.
- Write 5 blog posts. I find that I like writing about what I’m learning and it gives me an excuse to practice D3.js.
- Go abroad three times with Lise.
- Finish the remaining major renovation projects so the apartment is ready for the summer. Last summer was extremely hot and our small apartment felt like an oven. I ended up sleeping on our terrace one night.
- Get my mortgage down to 200k dkk.
I’m looking forward to next year.