GenAI / Applied AI Engineer building retrieval, evaluation, and agentic systems in Python on AWS — focused on making AI systems verifiably correct, not just impressive in a demo.
Currently: Data Science Intern @ Kotak Life Insurance — testing and evaluating a GenAI Sales Assistant, and building a RAG evaluation & benchmarking framework on AWS Bedrock (RAGAS, DeepEval).
📍 Bengaluru, India · 🎓 B.Tech CSE (AI & ML), JAIN University — 2026
A full-stack RAG system answering factual and comparative questions across 6 insurance products straight from PDF brochures — hybrid BGE-M3 dense + BM25 sparse retrieval (RRF), BGE cross-encoder reranking, and Qwen3 32B generation via AWS Bedrock, with a citation-verification step that checks every generated claim against retrieved source text before it's shown.
Scored: Faithfulness 0.936 · Context Recall@4 0.837 · Answer Correctness 0.725 (RAGAS, 30-question benchmark)
Python FastAPI Next.js Chroma AWS Bedrock
A LangGraph 4-agent pipeline (Planner → Investigation → Reflection Gate → Resolution) that proposes refund/replacement/escalation decisions — every proposal is validated by deterministic Python policy checks before execution. "LLM proposes, Python validates, Action executes."
Scored: 100% issue-classification · 93.94% resolution accuracy · 96.97% final-status accuracy (33-scenario suite, 2.29s avg latency)
Python LangGraph FastAPI Next.js Docker
A side-by-side demo exposing training-serving feature skew: a LightGBM fare model scored through duplicated ("Broken") vs. shared ("Correct") feature-transformation pipelines, across 9 features and 3 injectable skew scenarios.
Showed: a distance-unit skew silently shifting a predicted fare from ~$20.29 (9/9 parity) to ~$29.19 (8/9 matched) — despite a valid API response either way.
Python LightGBM FastAPI React pytest
A pipeline processing 7.08M U.S. domestic flights (BTS, full-year 2024) through PostgreSQL bulk-loading and tail-number-based aircraft-rotation reconstruction, independently estimating delay propagation between an aircraft's consecutive legs and benchmarking it against the BTS's official attribution.
Found: 95.28% valid aircraft-link rate · 0.764 buffer-adjusted correlation · downstream delay rate falls from 69.06% (0–30 min buffer) to 15.29% (120+ min)
Python PostgreSQL SQL Plotly Dash
GenAI & LLM: RAG · LangGraph · AWS Bedrock (Converse API, Bedrock Mantle) · Strands Agents · LLM evaluation (RAGAS, DeepEval) · Prompt optimization & token accounting Cloud (AWS): Bedrock · EC2 · S3 · SageMaker AI · IAM · DynamoDB ML & MLOps: LightGBM · scikit-learn · feature engineering · MLflow · Docker · GitHub Actions (CI/CD) · model monitoring Backend & Data: Python · SQL · FastAPI · Pydantic · PostgreSQL · SQLAlchemy · pandas · NumPy Visualization: Plotly Dash · Tableau · Power BI
