TRIAGE / SYSTEM01—03
TRACK CLASSIFICATION HELD-OUT F1 SCORE
82%
✳ AI JUDGE CO-PILOTOrganizer effort−80%
MACHINE LEARNING · NLP · HACKATHON TOOLING
Hackathon ML Triage & AI Judging System
Uses TF-IDF and Logistic Regression to classify project abstracts into technical tracks. A Groq / LLaMA 3.1 judge co-pilot generates architecture summaries and award justifications, while GitHub commit-history validation flags activity that predates the hackathon. Achieved 82% weighted F1 on the held-out test set and reduced estimated organizer triage effort by about 80%.
View on GitHub ↗