Led development of an AI-driven hiring platform (later spun out as a standalone startup) that conducts real-time LLM-driven avatar interviews and produces structured, multi-dimensional candidate analysis.

Real-time avatar interviews: low-latency conversational interviews with prompt-tuned LLM personas.

Multi-agent scoring: independent scoring agents evaluate communication, technical depth, and role fit, with a verifier agent cross-checking for consistency.

RAG-augmented evaluation: candidate responses grounded against role-specific knowledge bases.

Turned a multi-week manual screening process into an automated same-day pipeline with structured, comparable candidate dossiers for every applicant.

Python, Node.js, streaming conversational stack, LLM APIs, vector databases, React.