We all take the one and a half hour bus home every Monday and Friday. To spend the time we've taken up reading and have recently been really interested in the book Superforecasting by Tetlock. The premise of the book is that large groups of individuals can generally make very accurate predictions using the average. More than that, there are individuals who happen to do many levels better than the competition in predicting various events. Tetlock explores their habits, and we've also wanted to learn more about their skills and use them to their advantage.
What it does
Our platform connects top companies interested in high levels of insight with these superforecasters. The platform accesses and keeps track of these individuals allowing them to get compensated for their efforts. On top of that, we built in our own analytics to attempt to aid the superforecasters in their predictions, based on insight we learned in the book.
How we built it
We built the platform to have two many user types. It was written in react, using new methods for keeping state (undux) and navigation (navi). For the backend, we first began using firebase but had issues with scaling up the data so we built our own backend connecting Graphql to Prisma to PostGre. The analytics are a micro-service written in golang that communicates with google's cloud AI platform.
Challenges we ran into
The biggest issue we ran into by far was using typescript for the first time. While it is a powerful system good for scaling securely, in this competition setting and having to deal with the irritation of defining proptypes every time, our efficiency was greatly decreased. Other issues were firebase's ability to handle many gigabytes of data.
Accomplishments that we're proud of
We built a fully working backend and have stitched parts together with the front end. We committed to using typescript for at least 75% of the competition. We also sped up our communication with google's AI platform using some tricks.
What we learned
How to build sleek microservices, how to use typescript, making platforms, and, of course, teamwork.
What's next for Sybil
Deploying to the web.