ML Yield Prediction Model
with Dean Oya Tukel, Martin Tuchman School of Management
Universities guess at enrollment yield more than they admit. We're replacing the guess with a model.
It forecasts which admitted students will actually enroll, trained on 12,000+ records spanning
15 years of historical data. Current performance: 72% accuracy, 70% precision.
- My role: data cleaning, feature engineering, model training, evaluation
- Status: in progress — [SUMMER SYMPOSIUM SHOWCASE? Confirm]
- Methods: [FEATURE ENGINEERING DETAILS, MODEL SELECTION — needs Rishab's input]
Pythonscikit-learnpandasPredictive Modeling
AI Accessibility for Neurodivergent Users
with Prof. Andrew Klobucar, NJIT Humanities
Mainstream AI interfaces assume a neurotypical user. This research asks what changes when you stop
assuming that. We study how students with ADHD, anxiety, and autism actually use AI tools, and we
prototype interfaces designed for them from the start. This work feeds directly into AdaptIQ.
- My role: design research, user interviews, prototyping
- Status: in progress
- Research question: [EXACT FRAMING — needs Rishab's input]
HCIAccessibilityClaude APIPrototyping