Researched and built ML pipelines for extracting venue entrance locations from street view imagery and open datasets. Implemented GTFS based mechanism to identify entrances of 1500+ metro stations with near 100% accuracy.

Developed ML models to predict demand for 180+ food items from 15 features, reducing food waste in U-M dining halls. Built predictive interface for dining hall staff to interact with models, replacing former Excel-based calculations.

One of 10 student developers selected to build NYU's Classroom-LM RAG education chat platform using Gemini and a custom RAGFlow Engine. Built agent interaction interfaces using Next.js, TypeScript, and ShadCN, developed notes augmentation featureset to enhance and correct student text notes. Configured SQL tables within Supabase with Next server-actions paradigm for classroom management and realtime agent group chat functionality.

Developed ML paddling technique optimizer synthesizing 45+ Apple Watch motion data streams to provide comparison graphs, segmented analysis, and LLM feedback, leading to race speed improvements for 12-time regional champion team. Implemented G-Code optimization for steel CNC milling toolpaths, halving milling times and tripling success rate. Designed battery-electric chain-drive crate to transport 200+ lb canoe for competitions.

Constructed Unity models for machines like 3D printers in NYU’s Makerspace lab to allow remote training. Developed Meta Quest Oculus headset integration for interactive functionality on Tormach 770M CNC milling machine.