M.S Statistics & Data Science (SDS) @ IIT Kanpur | Data Science · AI/ML · Deep Learning · Statistics · Quant · Data Analytics
- 🎓 Pursuing M.S SDS, Department of Statistics & Data Science, IIT Kanpur
- 📊 I build statistically-grounded ML systems — not just models, but pipelines that hold up to validation: time-series diagnostics, backtests, and out-of-sample checks
- 💼 Interested in Data Science, AI/ML, Deep Learning, Statistics, Quantitative Analysis, and Data Analytics roles
- 🔭 Currently deepening my work in time-series econometrics and statistical machine learning
- 🌱 Always exploring how classical statistics (hypothesis testing, stochastic processes, econometrics) and modern ML complement each other
- 📫 Reach me at sumits25@iitk.ac.in or connect on LinkedIn
| Project | What it does | Stack |
|---|---|---|
| 🛡️ SpectraShield | Detects transaction fraud by uncovering a 24-hour spectral rhythm in the data (ADF/KPSS, Welch PSD) — XGBoost + SHAP, AUC 0.80 | Python XGBoost SHAP Streamlit |
| 🏦 Mortgage Credit Risk Modeling | PD/LGD/EAD credit risk framework on Freddie Mac loan data, validated against realized 2007-crisis losses | Logistic Regression XGBoost IFRS9 |
| 📈 RBI Sentiment & Bond Yields | Scores hawkish/dovish tone across 163 RBI policy documents with FinBERT and links it to bond yield moves (69% directional accuracy) | FinBERT Econometrics NLP |
| 🖼️ CNN for CIFAR-10 Classification | Deep CNN built from scratch in PyTorch to classify 60,000 images across 10 categories (74.71% test accuracy) | PyTorch Deep Learning CNN |
| 🍔 Big Mac Index Analysis | SQL + Power BI analysis of currency valuation across 57 countries using window functions | MySQL Power BI DAX |
| 📊 Black–Litterman Portfolio Optimization | Portfolio allocation framework using oil-price-shock-derived views, validated with walk-forward backtesting | Python Quant Finance |
| 🎯 Stock Market Anomaly Detection | Z-score and ARIMA-based anomaly detection for NSE stocks with a live R Shiny dashboard | R Shiny ARIMA |
🔎 More projects — resume screening, workout coaching, biometric attendance, and deep learning — on my repositories page.
Languages
ML / Data
Visualization / Apps
Tools
💬 Always happy to talk time series, credit risk modeling, or quant finance — feel free to reach out!