Mundo Raises $20M for Perceptual AI
Mundo raised $20M to build data and evaluation infrastructure for perceptual and multimodal AI.
Multimodal AI needs more than bigger models. Mundo’s round points to the data and evaluation layer required for systems that perceive images, video and real-world context more reliably.
What happened
Vancouver AI startup Mundo raised a $20M Series A led by GreatPoint, with participation from Y Combinator, Next Frontier and E12. The company builds data and evaluation infrastructure for perceptual and multimodal AI systems.
Its work sits around the question of how AI systems understand the real world, rather than only generating text or analysing structured data.
Why it matters
Perception is a bottleneck for robotics, autonomous systems, spatial intelligence and multimodal assistants. Models need high-quality datasets and evaluation loops to understand scenes, objects, movement and context.
That creates room for companies building infrastructure around data quality, benchmarking and feedback, especially as AI moves into physical and visual environments.
The bigger picture
The AI market is forming a deeper picks-and-shovels layer around model development. Compute is one part; data and evaluation are another. Mundo’s Series A shows investors backing companies that make models more grounded, measurable and useful for real-world perception tasks.