Ahtasham Athar

I’m a Senior Software Engineer with 3+ years building and scaling production full-stack and real-time systems, with a focus on voice AI that delivers reliably under live traffic. I architected backend infrastructure for a VAPI-style conversational-AI platform using async Python (FastAPI), bringing together multi-provider LLM/STT/TTS orchestration, function/tool calling, and RAG knowledge bases.

In my most impactful work, I built a production LiveKitAgents voice worker with real-time STT→LLM→TTS pipelines, ONNX semantic turn detection, and validated tool-calling for mid-call actions like lookups, bookings, and transactions. I also improved provider resiliency with automatic fallback across 8 LLM / 9 STT / 7 TTS providers, while achieving ~800–900 ms best-case and 1.0–1.5 s average end-to-end voice latency for live customers.

I also lead frontend delivery—owning Next.js/TypeScript architecture (Next.js 15, React 19, Zustand, React Query, NextAuth token-refresh queue) and shipping agent-builder, campaigns, analytics, and in-browser voice agent test experiences on the LiveKit web SDK. Earlier, I managed a team of 10–12 developers and designers, driving agile execution, runbooks/training, and measurable performance improvements.