At Tez Health, I built Dr. Tez, a multilingual medical triage voice assistant that handles voice and text in five Indian languages. I added real-time speech recognition, live captions, and safety gates for critical inputs.
I cut time-to-first-audio to about 1.5 seconds with a sentence-level streaming pipeline, async synthesis, and a Google TTS fallback. I also added latency instrumentation and a pytest suite for clinical scenarios and regressions.
In my projects, I developed TriageForge, a multi-agent ticket triage system with guardrails and an evaluation harness, and FloatChat AI, a conversational RAG backend for ARGO ocean profiles. I work with Python, FastAPI, LangChain, and tools including PostgreSQL and ChromaDB.
