Team
Fundamental Forecasters
Project Concept
Concept
Use betting market data from Kalshi to measure error margins and biases of different markets. In addion to this allowing for better risk estimation this provides an edge in betting.
Goals
Calculate expected calibration error across multiple market types. Check for upside and downside biases. Group markets and bets ( low original chance, political leanings, market type) .
Team Contribution
Helping with processing betting data, ideation on how to weight the timeseries, coding, dashboards etc.
Note: I have only done light research prior but this should be doable
Entry
Status: Submitted
Last saved: November 08 at 1:41 PM EST
Team Roster
Message board not available for this team yet.
Ethan Otto Team Lead RSVP Approved
ML Engineer at Armstrong World Industries
Data Scientist
MS CS Cornell
Coding agents
Dnd module creator
Scott Williams RSVP Approved
AI Engineer at Muse
Data Engineer
I'm Scott, a graduate of The Colorado School of Mines with a Bachelor's in Computer Science and focus in Applied Math. I interned last summer, and have accepted a return offer, on the Data Engineering team at iCapital in NYC. During college, I spent almost as much time studying statistics and computational applied math, as I did focusing on CS concepts. Naturally, these two interests of mine intersected well in the AI/ML space. Additionally, I have built many projects using AWS/GCP, so currently I would say my skills fall heavily under the AI infrastructure/Engineering umbrella. In addition to my technical interests, I also love reading historical books, and am a lifelong athlete, as I have been playing ice hockey since I was 6 years old.
My website: https://scottwilliams.win/
I am interested generally in the information retrieval and computer vision fields of study in the CS/AI/ML world.
I have always been interested in the finance space. Specifically, I find the activities of both fintech companies (such as my current employer, iCapital), and quant/automated-trading firms (such as Citadel, HRT, etc), to be fascinating.
Finally, I am interested, and have worked on projects, relating to automating content-creation for marketing purposes on social media platforms.
I am currently working on a tool called Muse with my co-founder, JP Broz. The tool uses fine-tuned vision embedding models, and a custom query engine, to efficiently search millions of rich document pages in an accurate manner. Once complete, the product will be an out of the box, end to end, solution for document retrieval across enterprise document collections. I am also working on optimizing for GPU utilization, to pass cost savings to end users of Muse.