// AI/ML Engineer & Researcher
Architecting intelligent systems through advanced Deep Learning and Generative AI.
Focused on bridging theoretical ML and scalable real-world automation.
I design and ship agentic AI systems — multi-agent pipelines, RAG architectures, and predictive ML models that go to production. Not demos. Tactical planning, high-velocity iteration: ship the MVP, break it, refine until it's right.
| # | Project | Stack | What it does |
|---|---|---|---|
| 01 | SHPSv2 | XGBoost · LSTM · SHAP · Flask · Docker | Predictive asset management — 95% CI structural longevity forecasting (R² 0.98) |
| 02 | Zenic | LangGraph · Llama 3.3 70B · Qdrant · Ragas | Agentic health AI — hybrid BM25 + vector RAG, mandatory safety gates, Faithfulness > 0.85 |
| 03 | SoundReverse | LangGraph · Gemini Flash · FastAPI · Pydantic v2 | Analyst-Critic pipeline that reverse-engineers mastering decisions from raw audio signal metrics |
| 04 | AR/VR Campus Nav | Next.js 16 · Mapbox GL · Zustand · Vercel | IEEE-published browser-native AR navigation — 4-mode FSM, no WebXR runtime |
| 05 | Stem-Split MCP | FastMCP · HTDemucs · CLAP · Librosa · Docker | Custom MCP server — decompose any YouTube track into 4 stems + 512-dim CLAP embedding |
Agentic LangGraph LangChain FastMCP Llama 3.3 70B Gemini
ML / Research XGBoost TensorFlow Keras PyTorch SHAP HTDemucs LAION CLAP
RAG & Eval Qdrant ChromaDB BM25 Cross-Encoders Ragas LangSmith
Backend FastAPI Flask Pydantic v2 Docker Vercel HuggingFace Spaces
Frontend Next.js 16 React 19 Tailwind v4 Zustand
IEEE ICISS 2025 — AR/VR based Campus Navigation System
Lead Researcher & Developer · Flutter · Mapbox GL · ARCore · Unity
→ DOI: 10.1109/ICISS63372.2025.11076255