Economics student building production-grade data science systems

I'm an MA Economics student who works at the intersection of economic reasoning and machine learning. My focus is on projects where the modeling decisions have real business or policy consequences — not just leaderboard scores.

End-to-end loan default prediction system built on 307,511 applicants across 8 relational tables.

• Model: LightGBM · AUC 0.783 · KS 42.7%
• Business layer: Threshold optimized against asymmetric costs (33:1 FN/FP ratio)
• Outcome: ₹93.9 crore net business value on held-out test set
• Stack: Python · LightGBM · FastAPI · scikit-learn · pandas

The interesting part wasn't the model — it was calibrating the decision threshold against actual loan economics and ensuring the feature pipeline runs identically in training and at inference time.

Macroeconomic time series modeling using Indian national accounts data. Applying econometric and ML forecasting methods — ARIMA, Prophet, and gradient boosting on lagged macro indicators.

• Degree: MA Economics
• Focus: Predictive modeling, feature engineering, business-driven ML
• Interested in: Data Scientist roles where economic intuition and ML intersect

My economics training shapes how I frame problems — cost asymmetry, incentive structures, and causal reasoning are things I think about naturally, not as add-ons.

Feel free to open an issue on any repo if you want to discuss the work.