NSF AI INVITE Institute
- Built a real-time learning analytics tool that helps AI tutors catch struggling students mid-lesson, capturing 9 live signals from student code and robot movement at sub-second latency.
- Shipped a live session dashboard, demoed to NSF site visit reviewers, turning 41K+ events into structured context for AI tutors.
- Cut database storage 84% with a normalized 5-table redesign, recovered 24 hours of silently dropped logs per query, and added a health endpoint that auto-degrades to 503.
- Launched an agent intervention platform letting a 10+ person team deploy tutor behaviors by config change, evaluating rules against 200K+ events on a 5 second cycle.