My projects turn statistical and machine learning methods into end-to-end systems that are tested against known ground truth. In SaaS Finance Analytics, I built an ASC 606 revenue recognition engine and close-reconciliation checks that caught a billing discrepancy with zero false positives.
I built an Experiment Bench with power analysis, CUPED variance reduction, and sequential testing, catching false-positive inflation before it could bias a ship decision. My Causal Inference Simulator brings together estimation, refutation testing, and heterogeneous treatment effects.
For a real estate data pipeline, I used PySpark and BigQuery to process large datasets and built an XGBoost model whose price coefficients matched ground truth. I also developed a graph-based fraud detection pipeline, comparing GraphSAGE, GCN, and GAT models and serving inference through FastAPI.
As a Data Science Intern at Step Computers Pvt Ltd, I built Python ETL pipelines that automated recurring stakeholder reporting. At I-tech Worx Private Limited, I automated data extraction and transformation and prepared datasets for downstream machine learning work.
