Vapi is a voice AI infrastructure platform founded in 2024 by Jordan Dearsley and his University of Waterloo classmate Nikhil Gupta. The company builds the orchestration, latency, and reliability layer that sits between large language models, speech-to-text engines, text-to-speech voices, and the telephony network — letting developers ship production-grade voice agents in days rather than months.
The origin story is that Dearsley built a voice-based AI therapist for his daily walks in 2023, chaining models together and hand-tuning latency until he had a working phone-based system. The duo pivoted that prototype into Vapi and launched publicly in 2024. By May 2026 the platform had handled over 1 billion calls and counted Amazon Ring as a marquee enterprise customer after Ring chose Vapi over 40 competing vendors.
Vapi raised a $50M Series B in May 2026 led by Peak XV Partners at a roughly $500M valuation, with participation from M12 (Microsoft's Venture Fund), Kleiner Perkins, Bessemer Venture Partners, and earlier investors. The round brought total funding to approximately $72M and followed reported 10x growth in enterprise ARR.
The platform is developer-first. It exposes REST and WebSocket APIs, a CLI for local development, SDKs for major languages, and plug-in support for swapping LLMs (OpenAI, Anthropic, Google, open models via Groq or Together), STT providers (Deepgram, AssemblyAI), and TTS voices (ElevenLabs, Cartesia, PlayHT). Built-in features include interruption handling, end-of-turn detection, function calling, knowledge base RAG, and call recording with compliance controls.
Vapi differentiates from end-to-end voice products like Bland and Retell by focusing on the infrastructure and orchestration layer, giving enterprises maximum control over which models, voices, and telephony providers they use. That positioning is why it has won large enterprise deals where reliability, observability, and compliance matter more than a turnkey UI.