I build voice AI systems that hold up in production. Speech models, real-time infrastructure, and the orchestration layer in between.
Founder of Mahimai AI Solutions — production voice AI on the open stack: LiveKit and Pipecat.
What I work on
- Fine tuning and evaluating TTS and STT models
- Real-time voice infrastructure: LiveKit, Pipecat, WebRTC, SIP
- Linguistic frontend, turn-taking, and interruption handling
- Concurrency, latency, and cost at production scale
Contributor to livekit-agents · FastRTC
Author of Voice Agents Handbook
|
How your voice travels through WebRTC Your smoke run is lying to you Building a phone line in PyTorch The normaliser that scored itself More on mahimai.ca/blog |
Add to onboarding reproduction logs feat: add livekit-plugins-sambanova with LLM support 🐛 Fix Feat: FastRTC version of Whisper CPP speech to text to Docs feat: Added documentation for twilio integration |
|
openrtc 0.20.1 voicegateway 0.26.1 envoic 0.3.1 locallens 0.0.2 fastrtc-whisper-cpp 0.1.2 |
Voice AI · repo Realtime Voice · repo Awesome TTS · repo Voice Prices · repo Voice AI Skills · repo |
| Package | What it does | Latest | Downloads |
|---|---|---|---|
| voicegateway | Observability and inference routing for voice AI | 0.26.1 Sep 2026 |
 |
| openrtc | Shared memory layer for multi-agent LiveKit workers | 0.20.1 Sep 2026 |
 |
| envoic | Horizontal voice AI infrastructure toolkit | 0.3.1 Jul 2026 |
 |
| fastrtc-whisper-cpp | whisper.cpp STT backend for FastRTC | 0.1.2 May 2025 |
 |
| fastrtc-canary | NVIDIA Canary STT backend for FastRTC | 0.0.2 Mar 2025 |
 |
| vapiserve | Custom tool server for Vapi | 0.0.5 Mar 2025 |
 |
| locallens | Local inference tooling | 0.0.2 Apr 2026 |
 |
Python, with Go where it matters.
Available for voice AI consulting. Book a call
The lists above rebuild themselves every six hours. How this works