I recently discovered a great podcast called The Startup Scientist, of which I have now swallowed up all episodes – thanks to Daniel Lakens for the recommendation! The podcast is about treating your academic career as a start-up. I have made this comparison before when giving talks at career orientations but wasn’t able to distill this idea, so I was excited to discover that Dan Quintana did exactly that, so now I can add on to it in this post.
The episode I want to write about is Reverse Engineering your Career Goals. Although I do often say that my career so far is a series of coincidences, I do think this type of reverse engineering has played a part in it.
Towards the second half of my PhD I had a lot of doubts about continuing in academia – I really wanted to, but I was aware that it would be very difficult to get a position. I wanted to do a postdoc, but only if I had a reasonable chance of getting a position afterwards. I was not sure how to estimate this chance, so I used the approach below.
1. Find representative data
I first studied the CVs of Dutch professors I knew, and concluded that most of them received independent funding a few years after their PhD. But these professors were already professors for a few years, so I decided their situations did not apply to me.
Through the funder’s website, I found a list of people in related disciplines who received this type of funding that year. Since I was a year away from finishing my PhD, and these people got their PhDs maximum 3 years ago, the gap between me and them narrowed.
2. Generalize
After searching for CVs of my “prototypes”, as a good machine learner I tried to find patterns. Although the people were all quite different, and I only had a few examples (I wasn’t doing this automatically, although that would have been an awesome side project), it was easy to spot several things they had in common.
I did this during my pre-Evernote era so sadly I do not have notes on what exactly I discovered, but I do remember two things in particular:
- All recipients had an impressive (at least for me) h-index. The minimum I found was 6, but the values between 9 and 12 or so were more common. This will vary per field, but for context, the professors I thought were not representative, would be in the 20+ category.
- Many recipients had an extra “thing” that was different from most others, like their own company, an organization they volunteered for, etc.
3. Multiple time points
Next to looking at people’s websites, it’s also helpful to search for “Name CV filetype:pdf”. Using this method, I was also able to find CVs of the same people, but from a few years ago. This had several benefits.
While the current CVs looked quite intimidating to me, the “time travel” CVs were much more relatable. Instead of learning only about the “value” of a CV, I was now learning about the “gradient”, which would be easier to apply to my own situation.
Of course, it is possible to do this for any CV, by removing all the things from the recent years. But the nice thing about the real “time travel” CVs was that I also saw how people changed the way they presented the same thing, that both CVs already had. For example, a side project that might have been insignificant in the “time travel” CV, was described in more detail in the current CV.
4. Predict?
Now that you’ve estimated your function where the inputs are the CVs at different time points and the output is receiving funding, you could try to fill in what’s missing from your future CV, to get the same output.
Don’t forget that there are multiple solutions (so different inputs will lead to the same output) and noise (so the same input might lead to a different output). Since I’m on a roll with this analogy – there are also lots of other inputs that you might not even be considering yet (life and stuff).
I did not actually create any concrete plans of what I was going to do, based on this informal study. But I suppose that the information got deposited somewhere, and helped me make choices that would later point me in the direction of a solution.