Mundo has closed a $20 million Series A, bringing total funding to $24 million including a previously unannounced $4 million seed round. The company builds datasets and evaluations used by AI labs to develop what it calls perceptual intelligence — the ability for AI systems to understand real-world sensory experiences including speech, video, gestures, and the implicit signals that characterize human interaction.

The distinction Mundo draws is between reasoning and perception. Current AI systems have become increasingly capable at reasoning over structured information, but much of human communication is unstructured — conversations overlap, tone and facial expression change meaning, people gesture instead of name things, and context shifts constantly. Mundo's position is that the next generation of AI systems will need to handle those signals, not just text.

The company's datasets and evaluations are designed to help researchers build and measure progress in these areas, covering natural speech-to-speech interaction, fine-grained video understanding, and modalities that don't yet have established learning methods. Mundo frames its role as part of a feedback loop between better data and better models — as AI systems improve, the benchmarks and datasets used to shape them need to evolve alongside them.

The Series A will be used to expand the team across research, engineering, and operations.