— 000m · Surface
Voice AI · CPaaS · SIP · WebRTC
Voice AI, built after hours.
Day job: Staff Team Lead, Voice Support Engineering at Twilio. Scaling support quality through tooling, documentation, and people development.
After hours, I'm building the kind of systems I support — voice agents on Conversation Relay, ElevenLabs against Deepgram on real calls, Claude wired into both. The expertise that only comes from shipping it yourself and watching it break.
See the work- May 17, 2026
iOS Safari audio sessions: fifteen commits to a working voice mode
What it took to make a continuous-mic voice agent reliable in Safari on iOS. AudioContext routing, mic prime-and-release, the W3C Audio Session API, and the failures that taught me each one.
- May 17, 2026
Pre-render your TTS audio, don't stream it
Why streaming every utterance is the wrong default for production voice apps — the cost shape, the quota wall, the latency tax, and the pre-render pipeline I built to replace it.
- May 17, 2026
Three deploy pivots: each fix bought time, not a solution
RecallIQ's deploy pipeline went from build-on-EC2 to build-on-Mac to GitHub Actions. Three forced pivots, three walls, three fixes that each looked permanent for a while. The pattern of right-sizing where the work actually happens.