Mundo AI raises $24 million to build data layer for perceptual intelligence

Vancouver-based Mundo AI has closed $20 million USD in Series A funding, adding to a previous $4 million seed round for a total raise of $24 million. The capital is earmarked for expanding the company’s team across research, engineering, and operations.

While current large language models (LLMs) have demonstrated remarkable capability in reasoning over structured information, solving math problems, and writing code, Mundo notes that understanding unstructured reality remains a significant challenge. The company aims to address this gap by providing specialized datasets and rigorous evaluations for multimodal systems involving audio, video, and other sensory inputs.

This pivot suggests that the industry is recognizing data scarcity not just in volume, but in type. Building high-quality, labeled, real-world perceptual datasets requires significant effort and capital, creating a new infrastructure play. For AI labs, this means multimodal capability will require continuous feedback loops of increasingly specialized benchmarks.

The investment from GreatPoint Ventures, alongside Y Combinator and E12 Ventures, underscores the market's interest in foundational data tooling rather than just application-layer models. The focus is on solving the input problem for future generations of AI.