Graduate Research Assistant — Performance anomaly detection and benchmarking (CMU)
Built an anomaly detection platform and investigator-ready dashboard for athletics performances to support anti-doping prioritisation. Implemented and validated multiple detection methods against real-world sanctions data to enable explainable flagging and operational decision queues. Delivered tooling to audit results and compare methods for performance anomaly identification. • Labeled/curated athlete performance data slices across 2010–2025 (19k+ events) for benchmarking and method evaluation • Produced explainable anomaly flags using statistical, ML, and Bayesian detection approaches • Validated outputs against Athletics Integrity Unit sanctions records for real-world calibration • Maintained audit trails and prioritisation queues for investigators’ review workflow