shivani kaeley

Developed an automated BSE filings analysis system expanding coverage from 60 to 200 companies and reducing processing time by ~20 minutes per report; engineered automation pipelines that reclaimed 10 hours weekly and established quarterly AI initiative prioritizing high-impact projects.

Architected an NL-to-SQL engine with GPT-4o, Athena, and Redis Vector DB achieving 98.4% accuracy and 5s p99 latency, saving ~5 hours per ad-hoc query; built text-to-image and personalized content pipelines that reduced turn-around-times from weeks/days to seconds.

Led data science initiatives including developing an XGBoost model that identified 4% additional users for credit-limit decreases and produced behavior models that improved salaried and student eligibility while lowering portfolio risk.

Built behavior models using logistic regression and random forests that increased salaried eligibility by 12% (with 40% lower portfolio risk) and student eligibility by 125% (with 22% lower risk); automated KYC features from device data to raise approval rates.

Supported empirical and geospatial research on political favoritism by creating and cleaning datasets and conducting analysis using regressions, PCA, and Double LASSO to inform variable selection.

Grade: 3.76/4.0

Activities and societies: Research Assistant at the Center for Effective Global Action working on empirical and geospatial analysis; coursework and projects in data science, machine learning, and economics.

Completed a Bachelor of Arts with a double major in Data Science and Economics, graduating with a 3.76 GPA and engaging in empirical and geospatial analysis for research projects.