AI-Powered Surgical Instrument Counting System (ECS 193 Senior Design)

Tech: PyTorch, Swift, REST APIs, PostgreSQL, CoreML

• Engineered end-to-end ML system detecting 30+ surgical instrument types, processing 600+ instruments/hour to eliminate counting errors in OR workflows.
• Built scalable RESTful API backend with PostgreSQL, implementing caching layer to serve 500+ concurrent users at <100ms latency for real-time inference.
• Developed and published iOS and Android app with CoreML model compression, enabling offline prediction and auto-syncing discrepancy alerts to hospital inventory systems.