Senior LiveKit Engineer (Self-Hosted Voice Infrastructure)

NeuroDrift is a US-based AI voice and enterprise software company. We build real-time voice AI agents that run live on real phone lines, at scale, for enterprise contact centres. Our platform (CallDash) carries production call traffic every day, so latency, telephony quirks and audio edge cases are our daily reality, not a research problem.

We run LiveKit self-hosted, end to end: LiveKit server (SFU), the LiveKit SIP service, Egress, and LiveKit Agents workers on Kubernetes. We are hiring an engineer who has run this stack, not one who has used it.

Who this is for

If your LiveKit experience is LiveKit Cloud plus the Agents quickstart, this is not the role. We need someone who has:

- ⁠ ⁠Deployed and operated LiveKit server outside LiveKit Cloud: config, Redis for multi-node, TURN/ICE, ports and firewalls, version upgrades
- ⁠ ⁠Configured the LiveKit SIP service against a real carrier trunk: inbound and outbound trunks, dispatch rules, SIP participant lifecycle, transfers
- ⁠ ⁠Built on the LiveKit Agents framework and understands the worker model: job dispatch, prewarm, load thresholds, graceful drain, and what happens to a live call on redeploy
- ⁠ ⁠Debugged a bad call across room events, participant tracks, SIP logs and agent logs, not just inside the agent process

What you'll do

- ⁠ ⁠Own our self-hosted LiveKit stack in production: SFU, SIP service, Egress, Agents workers
- ⁠ ⁠Build and operate voice agents end to end: telephony, audio pipeline, STT/TTS, LLM integration, tool calling
- ⁠ ⁠Own latency. Chase every millisecond from end-of-utterance to first audio, and know which component owns each one
- ⁠ ⁠Harden SIP trunks against real carriers and contact-centre platforms: transfers, header passthrough, call correlation




- ⁠ ⁠Run stateful agent workers on Kubernetes: capacity, autoscaling, graceful drain, deploys during live calls
- ⁠ ⁠Tune turn-taking. Endpointing, VAD, barge-in and interruption behaviour are where calls feel human or don't
- ⁠ ⁠Build the observability that makes a bad call explainable after the fact, not just reproducible
- ⁠ ⁠Join client calls, present your approach, and defend your technical decisions

What we need

- ⁠ ⁠Production LiveKit: self-hosted server, SIP service and Agents framework. This is the hard gate
- ⁠ ⁠3+ years Python, including async (asyncio, FastAPI or similar)
- ⁠ ⁠Real-time audio fundamentals: WebRTC, SIP, RTP
- ⁠ ⁠Kubernetes for stateful workloads, not stateless web services
- ⁠ ⁠Comfort debugging from logs and metrics in an workplace you can't attach a debugger to

Nice to have

- ⁠ ⁠A SIP trunk you shipped and then debugged under load (Telnyx, Twilio, Bandwidth, Vonage or similar)
- ⁠ ⁠Egress recording pipelines and their failure modes
- ⁠ ⁠Contributions to LiveKit repos or active in the LiveKit community
- ⁠ ⁠Speech vendor tuning: Deepgram, Whisper, Azure Speech, keyterm biasing, telephony-band audio
- ⁠ ⁠Contact-centre integration: Genesys, Five9, NICE, warm and cold transfer, SIP REFER, UUI header correlation
- ⁠ ⁠Noise suppression and audio enhancement in a real-time path
- ⁠ ⁠Cost awareness: per-minute STT/TTS billing, and what a stuck call costs

The setup

- ⁠ ⁠Fully remote
- ⁠ ⁠High intensity, 50 to 60 hours a week. Startup pace, not a 9-to-5
- ⁠ ⁠Working hours primarily IST, with availability for US client meetings

You'll thrive here if you like owning hard problems end to end, you're comfortable in front of clients, and "it works on my machine" isn't in your vocabulary.

📌 Senior LiveKit Engineer (India)
🏢 Neurodrift
📍 India