The Role
We’re hiring a senior full-stack engineer who lives and breathes real-time audio. You’ll build end-to-end features across web and edge backends, own streaming STT/TTS integrations, and push p95 latency down while keeping accuracy up. You’ll collaborate with Product, Design, and ML to ship polished, reliable experiences at scale.
What You’ll Do
- Design and ship low-latency, streaming features (WebSocket/SSE/WebRTC) for real-time transcription, summaries, and AI prompts.
- Integrate and optimize STT (OpenAI Realtime/Whisper, Deepgram, Google/Azure/AWS) and TTS (ElevenLabs, PlayHT, Amazon Polly, Azure Neural), including streaming output, barge-in, and endpointing.
- Build and harden APIs/services (FastAPI/Node) with observability (OpenTelemetry), rate limiting, retries, and back-pressure.
- Implement audio pipelines: VAD, diarization, Opus handling, chunking/partial results, and transcript alignment.
- Work with LLMs for real-time coaching: function/tool calling, structured outputs, guardrails, and prompt design/evals.
- Own performance: measure WER, p50/p95/p99 latency, startup/first-token time, MOS for TTS; profile and improve.
- Collaborate on UI in Next.js/React/Tailwind; ship reusable components that feel instant under load.
- Ensure privacy & security: PII handling, data retention controls, tenancy, OAuth/SSO, audit trails.
- Write meaningful tests (unit/integration/e2e, load tests for streaming) and review code with care.
You Might Be a Fit If You Have
- 5+ years building production SaaS; 3+ years with TypeScript + React/Next.js.
- Strong backend skills in Python (FastAPI) and/or Node.js (tRPC/Express).
- Hands-on production experience integrating streaming STT and TTS with strict latency budgets.
- Web real-time chops: WebRTC, Web Audio API, MediaStream, Opus/PCM, jitter buffers, STUN/TURN/ICE.
- Cloud/serverless: Vercel, Cloudflare Workers/Queues, Supabase (Auth/Postgres/pgvector), Redis.
- Practical LLM integration: tool calling, evals, prompt safety, and cost/perf trade-offs.
- Telemetry mindset: tracing/metrics/alerting (OpenTelemetry, Sentry, Prometheus/Grafana, Datadog).
Bonus Points
- LiveKit/Jitsi/Twilio, SFU design, or media servers (Janus, GStreamer, FFmpeg).
- Rust/Go for audio/streaming services; WebAssembly for on-client DSP.
- Vector DBs & RAG (pgvector, Pinecone, Weaviate), LangChain/LlamaIndex.
- Zoom/Meet/Teams marketplace apps, OAuth scopes, and webhook ingestion at scale.
- On-device ML (VAD/denoise) and GPU-accelerated inference.
- Startup 01 experience; shipped something many users rely on.
Our Stack (Today)
Front-end: Next.js, React, Tailwind, Zustand/Redux, shadcn/ui
Back-end: Python (FastAPI), Node.js (tRPC), WebSocket/SSE, REST
Infra: Supabase (Auth/Postgres/pgvector), Redis, Vercel, Cloudflare Workers/Queues/KV/R2
AI: OpenAI/Deepgram/Whisper + vendor mix for TTS/STT; prompt engineering & evals
How We Work
Perks & Benefits
- Competitive salary + meaningful equity.
- Flexible hours; generous PTO.
- Annual AI/cloud credit stipend.
- BYO laptop + upgrade support.
- Learning budget for conferences/courses.
