Senior GenAI Engineer at JPMorgan Chase & Co.

Architected multi-agent orchestration systems using LangChain and LangGraph, building autonomous investigation pipelines with tool chaining, memory retrieval, and conditional branching to reduce analyst triage effort by 32% across fraud operations. Designed and shipped production RAG pipelines over internal policy and transaction knowledge bases using LangChain, ChromaDB, and OpenAI embeddings with parent-child indexing, hybrid search, and retrieval re-ranking. Integrated RAGAS evaluation (faithfulness, answer relevancy, context precision/recall) to reduce hallucination rate by 22% across two iteration cycles. Built and maintained FastMCP servers for Jira, Bitbucket, and browser automation, addressing credential configuration gaps across GUI-launched AI sessions. Instrumented pipelines with LangSmith and Splunk HEC for trace-level latency, tool-call success rates, and retrieval quality to meet enterprise observability and audit requirements. Standardized AI-assisted SDLC practices by p

Python AI/ML Engineer at TCS / Stryker Corporation

Built a FastAPI model-serving layer exposing a trained XGBoost complication risk model via REST endpoint for real-time pre-surgical risk scoring with sub-200ms latency. Engineered 15+ clinical features from 200K+ orthopedic records, contributing to a 15% AUC lift over baseline logistic regression. Trained and evaluated Logistic Regression, Random Forest, and XGBoost on imbalanced clinical outcome data, improving minority-class recall by 10% via threshold optimization and SMOTE resampling. Containerized the FastAPI inference service with Docker and deployed on AWS EC2 with CI/CD validation across dev and staging. Delivered Tableau and Seaborn dashboards to help clinical stakeholders interpret risk scores and feature contributions for adoption in pre-surgical workflows.