Final-year Data Science & Economics student at FLAME University. I build ML systems at the intersection of causal inference, computer vision, and development economics, with a bias toward problems that matter outside the lab.
Currently: ICDAR CMMHWR 2026 competition · presenter at ACBES 2026 · undergraduate researcher at RuDRA Lab, IIT Bombay.
Medieval Handwriting Recognition — ICDAR 2026 CMMHWR
6th Place Winner: ViT encoder-decoder + custom 1k-token BPE tokenizer. Current best: 0.096 CER (5-model ensemble).
Agricultural Price Volatility & Financial Inclusion — accepted, ACBES 2026
OLS, quantile regression, Random Forests, SHAP, and Causal Forests across Indian districts.
KisanSense
District-level crop stress prediction 14 days out from Sentinel-2, NASA POWER, ERA5, and ICRISAT data. SHAP risk scores + Hindi/English LLM advisories designed for SMS/IVR.
LLM Failure Mode Explorer
Systematic tooling for probing LLM failure modes across prompts and conditions.
Python R SQL Stata C++ PyTorch Hugging Face scikit-learn SHAP GeoPandas CUDA Streamlit LaTeX