Tulika Das

I’m an AI Engineer who turns LLM capabilities into production-ready systems—especially agentic workflows, retrieval-augmented pipelines, and evaluation-driven prompt engineering. In my internships, I’ve focused on reducing latency/token waste, improving response consistency, and shipping end-to-end solutions.

At Tulu Health, I identified redundant RAG usage for greeting queries and proposed a lightweight classification layer, targeting 40–50% improvements in token wastage and response latency. I also built a phase-driven LangGraph multi-agent orchestration MVP for healthcare triage with emergency-priority routing, structured state transitions, and audit logging.

I’ve also delivered strong results across projects—improving assessment accuracy by 80% through a LangChain-based framework and raising evaluation speed/reliability by 40% with multi-metric scoring. Whether it’s a complaint triage multi-agent system or a ReAct research agent, I prioritize clarity, traceability, and rigorous testing.