Data scientist working where statistics meets engineering. I care less about benchmark scores and more about whether a model holds up as a decision: class imbalance, delayed and noisy labels, confounding, drift, and the auditor who eventually asks why.
- 🎓 M.S. Data Science at Worcester Polytechnic Institute
- 🏥 Software Developer Intern at Bolster Rx, building healthcare data pipelines: eligibility ingestion, probabilistic record matching, privacy-preserving reporting
- 🔬 Previously: crypto-to-equity sentiment research at WPI, ML internship at Avnir, production computer vision at Isourse
- 📫 Reach me at ananyashukla2512@gmail.com
| Project | What it shows |
|---|---|
| fraud-detection-mlops | Cost-sensitive, explainable fraud engine with delayed chargeback labels, drift monitoring, retraining, and audit governance |
| causal-ml-treatment-effect | True treatment effects via propensity scores, IPTW, doubly robust AIPW, matching, and uplift curves |
| irrational-exuberance-in-the-age-of-algorithms | Research pipeline testing whether crypto sentiment spills over into equity returns (Granger, SVAR, agent-based simulation) |
| black-scholes-greeks-dashboard | Option pricer built from scratch with finite-difference Greeks in an interactive dashboard |
Languages: Python · SQL · Java ML & stats: scikit-learn · PyTorch · statsmodels · SHAP · causal inference · time series Data & infra: pandas · NumPy · Spark · FastAPI · AWS · Git