Fast enough to ship. Sharp enough to scale.

Built an end-to-end AI voice screening platform — a real-time STT→LLM→TTS pipeline over LiveKit WebRTC with a pre-LLM intent classifier (30–40% cost cut, 95% latency drop), a Pipecat conversation layer for natural interruptions and human-like filler, and a LangGraph semantic search microservice for multi-vector candidate matching.

Worked on building a complete project from start to deployment using the MERN stack. Designed a scalable architecture and implemented efficient CI/CD pipelines to streamline development and deployment processes.