2020 coming to an end

The past two end-of-years I have wanted to write a bilan de l’année. Both years were incredibly exciting and I had a lot to look back on. In both cases I wrote the notes for myself but never published them on the site. I guess that, while it is both useful and healthy to keep a monitor of how things have gone, I am not so keen on going public on these topics.

2020 has been a particularly unusual year. I am not excited about what I did this year. But I have to admit that this year, despite the obvious difficulties, brought a welcome return to modes of work which agree with me and, frankly, a blistering level of productivity.

I spent my Winter writing my first academic paper since leaving academia (3 years previously). I wanted to raise the bar in medical AI. I was deeply frustrated by my experiences in the Berlin AI / Start-up scene, where I felt that personal networks seem to speak louder than actual technical skills. So I published a guide to building an AI-driven medical product from kick-off right through medical device certification.

In parallel to writing this guide I took some time out to learn C++ (about 1 month) and then interview with a number of top tech companies. I have to admit, I really loved spending some time coding again. The Google interview was probably the most fun. Having somebody who is actually good at coding essentially shoulder-surfing while I wrote my code brought me back to my early 20’s when I first learned extreme coding. I passed the technical exams but didn’t get hired – possibly Covid had something to do with that.

I actually spent the first four weeks of Covid-in-Europe living in Paris. I had hoped to rebuild my network there and was strongly considering moving. As it happened, events were cancelled and all I could do was to enjoy jogging in the park and meeting up with old friends. I got back to Berlin on one of the last flights, out of Paris, before the European borders were closed.

Covid, we all know, sucks. My wife and I almost certainly had it. Apart from the 2 weeks where I was properly ill, I spent a further 4 weeks getting completely out of breath every time I climbed a stairs. Happily 2020 in Berlin has been a relatively easy-ride compared to other places in the world.

As the world shut down all of my plans fell apart. I had been invited to give two keynotes, at pretty big conferences, and I was an invited speaker at a major immunology conference. One after another, these events were all cancelled. As became clear over time, large companies all implemented hiring freezes. I was told twice that I was the preferred candidate for a role, only for a company-wide hiring freeze to be implemented.

I was lucky enough to have spent considerable time in the Winter mentoring a project which was developing a predictive model for adverse events during pregnancy. That project ran into some hiring delays so I offered to spend two months, half-time, developing the model to a professional standard for them. Those two months saved my sanity when I couldn’t leave the apartment any more. I also (again) re-learned that I love to actually build stuff. I rewrote and professionalised the entire pipeline before performing a handover to the incoming team. Things went so well that they asked me to come back later in the summer and help out, in a senior role, with the product direction.

I officially made myself self-employed – which is a really big deal in Germany. My clients – and their repeat business – has made this possible. Berlin is a difficult city in which to build medical AI. It’s kind of a small pond. Much smaller than Paris or London. Not everybody here needs somebody like me full-time. So I’ve been trying to size my offerings so that whatever I cost my client, I personally will add 10x that value in terms of their company (if a start-up) or product (for established players) valuation. That said, if a client is not forward looking, or if they have been ripped-off in the past, then they need to be able to trust me before paying such an amount. I am deeply grateful to all of my clients who have shown me their trust this year.

Part of my self-employed offering is as a mentor. Given that I am only 38 sometimes it seems ludicrous that I mentor others. But I have, compared to others, a lot of experience. This mentorship offering seems to pay-off for both me and my mentees.

Towards the end of the year, I suddenly decided to write an article about regulating medical AI. I have been surprised by how well received this article has been. I have not even advertised the article yet, and a number of senior industry figures have reached out to me to thank me for it. Christian Johner was particularly helpful in letting me know a few holes (see the changes from version 2 to version 3 on the arXiv) and also in encouraging me that it is already good enough. Publishing such an article, when you are no longer a proper academic, turns out to be exceedingly difficult. Thankfully, it is currently under review at a well-regarded, if somewhat low impact-factor, journal.

I think that I am going to wrap-up this entry, ironically for me, on the topic of writing. I hated writing for most of my life. I actually failed English in school. And I never wanted to publish when I was an academic. This year, something changed. Firstly, having written this blog for a number of years I better understood the power of writing in forming your own ideas. And secondly, I realised that writing is an amazing way to influence others. I’ve been quiet on the blog for the past couple of months, this is because I was writing so much elsewhere. There is the article in the pharma trade journal. The above mentioned articles, one published and one under peer-review. Finally, there are two further articles which are written but I am not handling the submissions. That is 4 peer-reviewed articles in one 12 month period. My entire previous academic career led to only 3 peer-reviewed articles!


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.