Inspiration
For both students and professors, getting feedback on what happens in class can be a great way to improve the learning experience. However, tools like Rate My Professor only sample from the extremes (people who really enjoy or hate the class) and only focus on the class as a whole. What if there was a way the whole student community could share their opinions on much more facets of university life? To address this we created Yomi, a prediction market for classes, sports, intramural, and just about everything else happening on campus.
What it does
Like all prediction markets, Yomi tries to predict the outcome of future events by taking input from its users. Each user can place a bet on a future event. Instead of using money like most other prediction markets, users bet with clout. Clout is a symbol of trustworthiness. When users correctly predict the future, they are rewarded with more clout, and they can lose clout if they are incorrect. Users can unlock markers of prestige in the shop if they have enough clout, and there is a leaderboard page showing who has the most clout. Integrated into the home page is a top 5 and bottom 5 clout leaderboard at each university. Users start with 100 clout, can gamble up to all of their clout on a bet, and can go negative with clout if they lose enough. Even if a user has less than 10 clout, they can still wager up to 10 clout on events. Users can see the odds of events happening and the leaderboards while signed out, but must be signed in to an account in order to place bets and create events to bet on. Anyone with an account can create a betting line on almost anything. They simply title the event and submit it. Our backend analyzes it to make sure it is appropriate, not a duplicate, and then it uses data to create the odds which users can bet on. When the event happens, the backend can either use AI that is monitoring websites or the creator can manually tell the app what actually happened, and then the clout is distributed. Yomi digs in much deeper into students' data to give students and professors a deeper understanding of how the community as a whole is affected by things on campus. By providing a more representative sample, Yomi provides a more holistic idea of the student body that can help professors better plan lecture and review sessions.
How we built it
For the frontend, we used React, NextJS, and Tailwind. Our backend is completely Databricks. We have a project with multiple databases that our programs writes and reads SQL requests to that store account information and all current betting lines. We used the Genie AI agent in Databrick to run calculations on odds and to update them every 15 minutes based on constantly incoming data. We host our app on Databricks Apps. The Databricks ecosystem made it extremely convenient, since every aspect of our backend and hosting was on one platform.
Challenges we ran into
About halfway through the build, one of us accidentally left an agent running and used all of our Databricks credits. This led to our entire backend collapsing and since it was 1am, we couldn’t contact the Databricks team for more credits. This led to us spending an hour and a half scrambling to find other options for our backend. We explored the possibility of using other AI integrated backend systems, but ultimately settled for making a new account on Databricks and copying our existing database and connecting our existing code to that. While we stayed on Databricks, this taught our whole team (including 3 first-time hackers) a lot about the critical role backends play and how they actually function, from SQL queries to HTTP requests.
Accomplishments that we're proud of
In less than 24 hours of coding, we built a fully functional prediction market with a clean, user-friendly frontend and an effective backend. We believe Yomi provides a much clearer sense of feedback than other avenues. Because of the broad appeal of prediction markets, Yomi will attract a greater variety and more representative sample of users than traditional feedback tools. Furthermore, because no real money is involved, there is no incentive to insider trade or sway the market and thus, Yomi will provide much more reliable data when it comes to how the campus community feels about future events.
What we learned
We all learned a lot about the many jobs a backend has to do. For most of the team, it was our first time working with SQL, HTTP, and APIs. It was also our first time integrating an AI agent into the backend with Genie. We also worked a lot with design principles while creating the frontend, focusing on it being visually pleasant while easy to navigate.
What's next for Yomi
We plan to keep developing Yomi with the goal of launching it. While developing, we conducted several interviews with the target audience and found a high level of interest in Yomi, while also receiving good feedback from potential users. After launching at UW–Madison, we plan to expand to other schools in the Big 10 and implement parlays, which will increase the appeal and user experience of Yomi.
Built With
- api
- databricks
- next.js
- python
- react
- sql
- tailwind
- typescript