Core Technology & Research
AI-Assisted Protocol Scraper & Database
We built a custom AI pipeline to scrape and structure historical scoring data from unstructured PDF protocols. This database indexes over 10 million individual scores and evaluation panels from over 25000 performances.
Behavioral Analyses of Judging and Technical Panels
Using structured competition results from 2022-2026, we completed the largest study ever conducted on judging panel and technical panel behavior. Our analyses highlight multiple ways in which bias manifests, who does it, who the primary beneficiaries are, and how it systematically affects skater placements.
Algorithmic Determination of Component Scores
We built computer models of each figure skating discipline to understand the relationship between component scores and executed element scores. We identified the main technical drivers of TCS and our models predict TCS more accurately than most ISU judges.
Kinematic Telemetry & Pose Estimation
Developing deep-learning computer vision pipelines to extract 3D kinematic telemetry directly from standard broadcast video feeds. By mapping pose keypoints to rink homography coordinates, our R&D framework derives objective metrics for skater velocity, jump height, flight distance, spin rotational speed, and team synchronicity without physical wearables.