An AlphaZero-inspired wargame simulating the historical Anglo-Mysore Wars, engineered for 100% in-browser execution with zero backend inference cost:

• State Compression: Compact 148-bit boolean state vector across a 959-action space.
• Client Inference: PyTorch self-play networks exported and INT8-quantized to ONNX WebAssembly, executing Monte Carlo Tree Search directly on the client CPU.
• P2P Multiplayer: Low-latency WebRTC data channels via PeerJS with deterministic state sync.
  _{PyTorch ONNX WebAssembly MCTS WebRTC FastAPI}

Diagnostic vision and clinical LLM pipeline developed during a Mayo Clinic research mentorship:

• Vision: YOLOv8 for sub-centimeter lung nodule detection in CT imaging.

• Language: Multi-stage retrieval-augmented generation (RAG) grounding Claude in verified clinical oncology literature, eliminating generative hallucinations.
  _{Python YOLOv8 Claude API RAG}

• Core Languages:
  Python • C++ • TypeScript • Java • WebAssembly • SQL

• Machine Learning & AI:
  PyTorch • ONNX Runtime • YOLOv8 • MCTS • RAG

• Web & Infrastructure:
  Next.js • React • FastAPI • AWS (ECS/Cognito) • Docker

• High-throughput self-play reinforcement learning pipelines.
• Low-latency client-side AI deployment using WebAssembly SIMD.
• Distributed systems and database internals.

• Actively coding and building projects on GitHub. Check out my pinned repositories above.

</td>