i'm priyadarshini, an analyst who likes crunching numbers and logical reasoning

Client capstone for Gaiadhi. Led a 4-person team building two LightGBM models in Python that classify cotton growth stage and warn 7 days before the fertilizer-critical stage, using 135K+ simulation records. Removed stage-proxy variables that inflated early scores, validated on unseen runs, treatments and future years, and engineered lag and rolling features. Delivered a phased deployment roadmap.

Estimated price elasticity across 3 grocery categories from 10M+ weekly scanner records (93 stores) with store-SKU and week fixed-effects regressions in Python: analgesics inelastic (-0.75), cereals and frozen juices more price-sensitive (-1.4 and -1.7). Modeled that a 5% mid-tier cereal price increase would raise annual gross profit ~19% ($580K) while cutting volume ~8%. Price-reduction promotions lifted gross profit ~113% vs. each store-SKU's regular-price weeks; one promo code (G) lost money.

Built ETS, ARIMA, ARIMAX and VAR models in R on 10 years of monthly interest rate data. VAR(2) cut RMSE 77% versus the ETS benchmark on a 25-month holdout, as its joint modeling of rates, VIX, TED spread, EFFR and exchange rate had the lowest error. ARIMAX predictors were selected by AICc.