Debrupa Behera

I've built end-to-end machine learning and AI projects that address telecom retention, customer targeting, policy support, and business analytics.

For telecom churn prediction, I used XGBoost, SMOTE, feature engineering, and SHAP analysis to achieve 82% accuracy, 81% F1, and 0.88 ROC-AUC. I deployed the solution as a Streamlit dashboard and FastAPI service on AWS EKS with Docker, CI/CD, and drift monitoring.

I also built a RAG insurance-policy chatbot with LangChain, OpenAI GPT-3.5, FAISS, citations, prompt-injection and PII guardrails, and multi-turn memory. My LangGraph Text-to-SQL analytics agent uses validation, self-healing query execution, role-based access, audit logging, anomaly scanning, and chart generation to answer business questions in under 10 seconds.

Backed by an M.A. in Statistics and training from DV Analytics, I bring statistical rigor to Python, SQL, machine learning, generative AI, agentic AI, and MLOps work.