Built production-grade AI applications (Python, FastAPI, Node.js, Next.js) across fintech, healthcare, operations, and knowledge management, reducing manual effort by 40–60%.

Designed RAG pipelines using LangChain, LangGraph, and pgvector/Pinecone, achieving sub-second retrieval and ~30% improvement in search relevance.

Developed low-latency LLM APIs (<300ms) with Redis caching, JWT authentication, and async processing, improving system reliability by ~25%.

Architected agentic multi-agent systems (planner–executor, doer–reviewer) using LangGraph with shared state management and tool orchestration.

Implemented checkpointing and human-in-the-loop workflows to ensure safe execution of critical actions and improve fault tolerance.

Designed conditional routing, retries, and failure handling in agent workflows, increasing end-to-end task completion rates.

Integrated external tools and APIs with scoped access control (read/write separation), improving system safety and preventing unintended actions.

Developed and migrated a monolithic pricing platform into an AI-driven full-stack microservices architecture on AWS EC2 using Node.js, NestJS, Python, and FastAPI, enabling real-time dynamic pricing at scale.

Served as a core developer, achieving optimal price calculations in under 200ms, significantly boosting market responsiveness.

Enabled AI-driven dynamic pricing decisions by building real-time data pipelines and distributed services that processed competitor and market signals at scale, supporting automated decision workflows across the platform.

Built a production-grade RAG chatbot for vendor and pricing intelligence documents using LangChain, pgvector, and semantic retrieval pipelines, enabling low-latency access to operational insights and reducing manual document investigation time.

Implemented caching solutions with Redis and in-memory Pod storage to accelerate data retrieval and reduce database load.

Enhanced database performance by 30–40% through advanced query optimization, strategic index creation, and automated purge mechanisms, resulting in faster response times and substantial storage savings.

Built a logging system integrated with Dynatrace for monitoring and diagnostics, reducing error detection time by 50%.

Performed code analysis, bug fixes, and functionality enhancements, improving stability and reducing error rates by 30%.