Convai

Convai occupies the embodiment layer of the AI-character stack: founded in 2022 by Purnendu Mukherjee, a former senior deep-learning engineer in NVIDIA's conversational-AI group, it builds the perception-dialogue-action engine that lets 3D characters in Unity, Unreal, and PlayCanvas see, hear, converse in 65+ languages, and act inside their worlds. Its NVIDIA relationship is the strategic asset few startups can replicate: a prominent Inception member and reference conversational layer for NVIDIA ACE, repeatedly showcased in NVIDIA keynotes and demos — including Jensen Huang's Computex spotlight — which positions Convai as the default "brain" vendor whenever the ACE digital-human stack lands a customer. The financial base is more substantial than the seed-stage image suggests: beyond the $5M 2022 seed from Monta Vista Capital and Rainfall Ventures, data providers report a roughly $10M Series A in late 2023 and third-party estimates put 2024 revenue in the mid-single-digit millions — meaningful commercial traction for a ~40-person company in a nascent category. The market it is early to is enormous on two fronts: games, where AI NPCs promise the biggest leap in interactivity since 3D graphics, and enterprise digital humans — brand ambassadors, training simulations, kiosk agents — where Convai's engine-agnostic, no-code-plus-API approach serves customers who will never build character AI themselves. As LLM latency and cost collapse, every barrier to shipping conversational characters at scale weakens, and the company that spent four years solving embodiment specifics — spatial perception, gesture, action APIs, real-time voice — inherits the demand.