I'm Pedro, founder of Fonoster: infrastructure for Voice AI. We run our own product on it, collecting real money on real phone calls. I write about what survives.
Free. One email per week. Unsubscribe anytime.
Infrastructure for Voice AI. Programmable voice from the carrier edge to the LLM, built for production from day one.
Our own product, our own traffic. Every lesson I publish was paid for here first.
For 15 years I've worked at the edge of communications infrastructure: SIP, RTP, media servers, and now Voice AI. I built Routr and Fonoster because the voice stack deserved better tooling.
But infrastructure claims are cheap. So we became our own customer: Qcobro, our vertical AI for collections, runs on Fonoster in production. Real calls, real money, real failure modes. That's where everything I publish comes from.
The goal is simple: make Fonoster the standard for how Voice AI gets built.
More about meFrom the carrier edge to LLMs. SIP, media, ASR, TTS, and agent orchestration. What they do, what they don't, and how to pick.
I take a small number of engagements a year, for teams building serious voice infrastructure.
I join your team to architect and de-risk your voice infrastructure, as the person who's actually built one at scale.
Three sessions. You walk away with a documented plan for what to build, what to skip, and in what order.
The missing manual for running Routr in production. Topology patterns, scaling, observability, and the failure modes nobody writes about.
At millions of calls a day, the cracks in the classic voice stack show, and why I'm choosing less in 2026.
Read essayMost Voice AI demos are theater. Here's what actually matters when a real customer answers.
Read essayA SIP server should be invisible. The lessons from 4 years of building one.
Read essay