HyrecruitAI
Full-timeCTO & Co-founder
2024 - Present·India · Remote
Leading engineering for an AI-powered interview platform. The job covers the whole stack — voice pipelines, LLM evaluation, multi-tenant SaaS, billing, and the team that ships it all.
01 · Owned
- →Entire engineering org and the technical roadmap
- →AI product surface — voice agent, evaluation, guardrails
- →Infrastructure — multi-tenant SaaS, observability, on-call
- →Hiring and engineering process — RFCs, code review, postmortems
02 · Systems
- →Real-time voice agent (Whisper + WebSocket + LLM turn-taking)
- →LLM evaluation engine with rubric pipelines
- →Multi-layer guardrails between LLM and candidate UI
- →pgvector + Redis semantic cache for LLM inference
- →Embedding-based candidate-to-job matching with reranker
- →WebRTC video stack with TURN/STUN and recording
- →Multi-tenant data isolation and per-tenant configuration
03 · Constraints
- →Sub-second voice turn latency or candidates drop the call
- →Zero tolerance for bias or PII leaks in candidate-facing AI
- →LLM costs that scale linearly with active interview load
- →Tenant isolation strong enough for enterprise procurement
04 · Impact
- →Scaled the engineering team from 2 to 15 engineers
- →Cut LLM inference cost 58% via two-layer semantic caching
- →Lifted interview completion rate from 34% to 71% after shipping voice
- →Brought top-10 candidate match precision from 34% to 81%
