IIT Madras (2nd Year)
I build machine learning systems focused on fintech, scientific discovery, fairness, and real-world decision making from finlife-os, to radiomics pipelines on cardiac CT imaging, to autonomous research agents, to fairness-audited policy rankers.
- 🔬 Scientific Machine Learning
- 🧠 Representation Learning & Clustering
- 🏥 AI for Healthcare & Medical Imaging
- 🤖 Autonomous Research Agents
- ⚖️ Fairness & Responsible AI
| Project | What it does | Stack | |
|---|---|---|---|
| 🛰️ | Antariksh Drishti — ISRO Bhartiya Antariksh Hackathon 2026, National Finalist (Top 30 of 60,000+ teams) | Five-stage exoplanet detection pipeline — BLS period search, physics-based transit classification, trapezoid fitting on real TESS light curves | Python · Streamlit · NumPy · SciPy |
| 🛡️ | RealityCheck | 6-agent system detecting scams, misinformation & undisclosed ads across text/URL/image/audio, with India-specific UPI fraud and Hindi/Hinglish handling | Python · LangGraph · FAISS · Groq/Anthropic |
| 🧠 | FinLife OS · | 6-agent Monte Carlo engine running 300 life-path simulations across 30 years SEBI-compliant, tax-optimized, shock-tested | Python · Streamlit · NumPy · Gemini API |
| 🔬 | Calcium Phenotype Discovery · | Cardiac CT radiomics pipeline Agatston scoring → PyRadiomics extraction → HDBSCAN clustering → UMAP visualization | Python · PyRadiomics · SimpleITK · HDBSCAN |
| ⭐ | AutoScholar | Autonomous research agent processes 100+ papers, extracts 8+ themes via BERTopic, evaluated with NDCG & MAP | Python · BERTopic · UMAP · Sentence-Transformers |
| ⚖️ | Policy Recommender AI | Fairness-aware ML ranker audited across 1,000+ simulated users Precision@K & NDCG across demographic groups | Python · scikit-learn · Fairness ML |
| 📈 | Marketing Conversion Prediction | Conversion model on 10,000+ campaign records identified 3 high-signal behavioural segments with feature attribution | Python · Pandas · Scikit-learn · Jupyter |
| 🎓 | Applied ML Case Study | End-to-end supervised ML pipeline with reproducible train/val/test splits IIT Kharagpur (KDH) coursework | Python · scikit-learn |
"The goal is to turn data into information, and information into insight."