Anjali Patel

I’m a Data Scientist/ML Engineer with production experience shipping credit-risk models, building end-to-end ML pipelines, and integrating analytics directly into decision workflows. At KPMG Global Services (KGS), I engineered XGBoost and LightGBM models on 500K+ loan records, lifting AUC from 0.76 → 0.89 while enabling an auto-approval system processing ₹2+ Cr in monthly disbursements.

I focus on performance, reliability, and explainability. I tuned PD cutoff from 40% → 35% using precision-recall trade-off analysis—recovering ~₹2 Cr in annual interest revenue from creditworthy borrowers previously rejected—while maintaining AUC-ROC at 0.89. I also implemented SHAP-based feature attribution for 100+ monthly underwriting reviews and deployed ML scoring via FastAPI REST endpoints with p95 <100ms, including health-check and retry logic.

Alongside traditional modeling, I build AI-driven systems such as LangChain/LangGraph agentic workflows and RAG pipelines. I created a LangChain + GPT-4 intelligent loan assistant that uses multi-tool reasoning, vector-database context injection, and LLM evaluation patterns across 100+ test scenarios—then packaged it with Streamlit and Docker for cloud-ready deployment.