Blog

Thoughts on voice AI, engineering, and building in public.

Engineering

Designing hallucination test suites for production voice agents

How we built automated testing that catches risky responses before your users ever hear them.

Infrastructure

Sub-50ms voice: architecture behind our edge network

A deep dive into streaming pipelines, model routing, and regional inference clusters.

Product

Multilingual agents without the complexity

One agent, 100+ languages — how VAPI handles locale detection, routing, and fallback.

Product

A/B testing voice: prompts, voices, and flows

Continuous optimization strategies for teams running experiments at scale.