My Summer with AI
Written by Alex Korman, Developer, Junior
Around December, I was aimlessly scrolling through Linkedin, half paying attention to some life science lecture. Bored. So bored.
I saw a post from a biotech AI company building “knowledge graphs”. The gist was: it’s impossible to keep up with the insane volume of biomedical research papers, clinical trials, FDA adverse events reports, and databases that are published everyday. Anyone who’s ever read a scientific paper knows it’s really hard to read just one, but navigating through the abundance of dense, novel information is near impossible. This company made a platform that dynamically augments search results by understanding the nodes a user adds to their knowledge map (like specific drugs, metabolic pathways, genes, etc.) in order to facilitate the discovery process for doctors, clinicians, and researchers.
Pretty neat.
I slid into the CEO’s Linkedin DMs with my resume and a short fan letter. I didn’t get a response, but I didn’t really expect to get one either.
HOWEVER, four months later pcv (post-coronavirus) he finally messaged me back out of the blue, asked when I was free for a call, and grilled me for an hour-plus about my data science experience.
That’s how I ended up at Epistemic AI for the summer working on an experimental project for their platform. My job was to see if I could create a few shot learning classifier for biomedical research papers.
Few shot learning is the practice of developing a reliable algorithm from a minimalist dataset. In this case, something that could read just a few paper abstracts, and extrapolate what it learned to thousands of others. The idea was that firms have their own internal labeling systems that they might want to apply to a corpus of documents.
I spent the summer reading up on natural language processing (which is how computers understand human language), emulating research and code I’d seen online, and trying out different techniques, network architectures, and machine learning algorithms.
My coworkers were super helpful and knowledgeable. It was really fun sharing the wins and losses with friends, while also working on a unique problem.
The fast paced learning environment was very similar to the one I experienced with ChangePlusPlus.
Generally at the start of the year, no one on a team is familiar with all of the technical requirements needed to build an application. You collaborate with your teammates to get up to speed, learn proper development techniques, and actually build something a client will use.
Knowing how to read online documentation, proper version control techniques with github, and work quickly was essential to my job.
CPP is a great playground to get acquainted with the ins and outs of product management and development with other people who really care about doing cool work. I’m close with my whole team and even collaborate with some of them on projects outside of school. It’s been a stimulating, educational experience that prepared me well for my summer internship.
