Lessons Learned During Insight

written by Eric J. Ma on 2017-07-17

(a) Solving healthcare goes beyond solving the science underlying it.

At its core, healthcare delivery is essentially a human problem. Even what we choose to optimize for is a hard problem. Do we optimize for changing human behaviour, or do we optimize for more precise treatments?

(b) Healthcare is complex

The biggest thing preventing a "solving of healthcare" is misaligned incentives.

(c) I like scientific data

Regardless of the lesson that healthcare needs to be solved with more than science, I still found myself naturally much more engaged with companies that were dealing with scientific data as part of their data science problems. Teams that were dealing with other types of data: insurance claims, financial, marketing, platform product analytics, click streams... these were much less engaging. I know my best fit now, though I won’t rule out other teams.

(d) People can change the equation.

I met with some people whose intellect and grasp of knowledge I really admire! Additionally, passion is infectious. It helps to work with colleagues who energize one another, rather than drain each others’ energy.

(e) Some Insight alumni are awesome

And I want to be like them when I help with mentoring for the next batch. Perhaps if I get a chance to interview others, I’d like to be able to model how I interview after the alumni mentors.

Biggest shout-out to George Leung, who works for Vectra, tailored his mentoring session by first asking me about my Insight project, which involved Gaussian processes and variational auto-encoders (VAEs). George asked me first about what VAEs were, and then asked me to solve a Bayes problem on the board. I could tell he was building his questions on-the-fly.

The other shout-out goes to Ramsey Kamar, who went through the “Big 4” questions: tell me about yourself, what’s your previous accomplishments, how did you face a conflict, and what’s your biggest weakness. His feedback to me was direct, positive, and most importantly, always encouraging.

(f) Humanities tools are needed

On reflection, I think that if we’re going to solve the “human” portion of healthcare, we’re going to need tools from the humanities - the tools that let us qualitatively and quantitatively study human behaviour. While data science can provide a quantitative path towards a solution, the qualitative side of it will remain as important as ever.