Data Scientist & ML Engineer · Research Intern @ NIT Trichy · IIIT Kottayam '27
I build end-to-end ML systems — from messy raw data through feature engineering, modeling, and production deployment. My work spans churn prediction, fraud detection, RAG systems, time-series forecasting, and computer vision research.
| Project | What it does | Key Results | Stack |
|---|---|---|---|
| Customer Retention & Churn System | End-to-end churn + LTV platform on 541K+ transactions with real-time scoring and NL analytics | AUC 0.82 · R² 0.99 · 16.87× ROI · £1.89M risk flagged | XGBoost · LangGraph · FAISS · FastAPI |
| RepoRAG | Production RAG over live GitHub issues — hybrid BM25 + Qdrant retrieval, cross-encoder reranking, RAGAS eval | Top-5 Precision ~80% → 100% | Qdrant · BM25 · RRF · Docker · FastAPI |
| Wind Turbine Forecasting | SCADA time-series pipeline for industrial asset performance monitoring and power output prediction | R² ≈ 0.97 | TensorFlow · LSTM · Scikit-learn |
| VICTOR (research) | Real-time multi-camera violence detection with identity-aware tracking and aggressor attribution | F1 0.8513 · 91.25% recall · 60× lower latency · 169K params | PyTorch · BiLSTM · OpenCV |
Languages
ML & Deep Learning
Generative AI & LLMs
MLOps & Deployment
- 🔬 Research Intern @ NIT Trichy — computer vision & temporal modeling
- 📄 Paper under peer review: "VICTOR: Identity-Centric Temporal Modeling for Multi-Camera Violence Detection and Aggressor Attribution"
- 🎓 B.Tech CS (Data Science & AI) @ IIIT Kottayam · graduating 2027
- 🏆 Smart India Hackathon qualifier · 200+ problems on LeetCode & HackerRank
