ABOUT IRFAN THANVEER
AI Engineer at Infosys BPM's AI Centre of Excellence, specializing in scalable multi-agent systems, deterministic retrieval architectures, and production LLM platforms.
Irfan Thanveer
Engineering Practical, Enterprise-Grade Autonomy
My engineering focus centers on moving AI systems from fragile conversational chatbots into robust, stateful, and deterministic enterprise agents. At Infosys BPM's AI Centre of Excellence, I lead and architect intelligent agentic solutions that handle complex reasoning, multi-turn state persistence, and autonomous tool usage.
I led a team of 3 engineers to achieve 1st Place at the Infosys BPM Internal AI Hackathon (competing against 20+ teams), architecting a LangGraph multi-agent intelligence platform that converts natural language business inquiries into verified analytical pipelines, automated data transformations, and publication-ready charts.
Prior to my focus on Agentic AI, I developed scalable cloud ETL pipelines and data warehouses utilizing AWS (Redshift, S3, Athena) and Python. This dual expertise in distributed data infrastructure and state-of-the-art LLM architectures gives me a unique advantage in building systems that are not just conceptually novel, but operationally reliable at enterprise scale.
CORE ENGINEERING PRINCIPLES
Standard vector retrieval frequently hallucinates and detaches table rows from structured figures. I design page-indexed, deterministic retrieval architectures that maintain 100% precision for compliance, finance, and mission-critical workflows.
Linear chains break under unexpected edge cases. By utilizing LangGraph StateGraph models with conditional branch routing, human-in-the-loop validation, and persistent memory checkpoints, agents gracefully self-correct errors.
Static scrapers fail when DOM structures evolve. I implement vision-guided autonomous browser agents via Browser-Use and Playwright running on Azure AI Foundry that intelligently navigate authentication, pagination, and dynamic interfaces.
Strict JSON schema enforcement (OpenAI Responses API & Pydantic), deterministic guardrails, secret protection, and observable state telemetry ensure enterprise data privacy and compliance at all execution stages.
READY TO BUILD NEXT-GEN AGENTIC SYSTEMS?
Whether you are exploring multi-agent workflows, vectorless RAG architectures, or enterprise AI consulting, let's connect and engineer something extraordinary.