Rohit Mishra — AI/ML Engineer@rht21

Delhi, India · Full-time

Building production LLM and real-time voice systems for AI-driven hiring — from proctoring and structured evaluation to LiveKit voice agents, with evaluation and reliability baked in.

• Built an AI video proctoring system for recorded interviews — detecting head-pose changes, eye-visibility issues, low motion, and exposure anomalies — cutting analysis time by ~80%
• Designed multiprocessing-safe video pipelines, resolving performance bottlenecks and resource contention during frame-level analysis
• Developed an AI interview evaluation system with structured scoring rubrics and schema-enforced outputs, reducing feedback turnaround by ~70%
• Built an end-to-end AI hiring pipeline — job extraction, requirement enrichment, screening-question generation, and candidate scoring — cutting setup time by ~70–75%
• Integrated LangSmith-based LLM evaluation with rubric-driven graders, regression testing across prompt/model changes, and cost/latency monitoring
• Built real-time LiveKit voice agents coordinating streaming ASR → LLM → TTS with sub-second latency, interruption handling, and stateful sessions

• Python
• LangChain
• LiveKit
• FastAPI
• OpenAI
• AWS