Ванкуверская Mundo AI получила 20 млн долларов на создание слоя данных для перцептивного интеллекта

Vancouver-based Mundo AI has closed $20 million USD ($28 million CAD) in Series A funding to build “the data layer for perceptual intelligence.” The news: Mundo develops the data, evaluations, and applied research that other AI labs and companies need to train systems capable of processing sensory information (or what it calls “perceptual intelligence”). The startup announced this financing last week, alongside a previously undisclosed $4-million USD seed round that brings its total funding to $24 million. Mundo’s Series A was led by GreatPoint Ventures, with support from fellow Silicon Valley-based investors Y Combinator (YC) and E12 Ventures, as well as Montana’s Next Frontier Capital. Mundo plans to use this funding to expand its team with hires in research, engineering, and operations. From the source: In its blog post, Mundo described perception as both “the missing half” and “the next frontier” in AI. The startup highlighted that existing AI models trained on data pulled from the internet “are becoming remarkably capable at reasoning over structured information” to write software, solve math problems, answer complex questions, and complete difficult tasks. But Mundo said another key challenge remains: understanding the physical world, which largely “remains unstructured.” The context: Mundo aims to help address this problem by providing the datasets and evaluations AI labs need to build the next generation of multimodal systems for audio, video, and more. According to its YC profile, the 30-person AI startup, which was founded in 2024 by CEO Jason Liao, COO Naijide Anwaer, CPO Garreth Lee, and CTO Kenneth Wu, initially set out to build “the world’s largest and highest-quality multilingual data library” to make up for the lack of AI training data beyond the English language. Final thought: These days, Mundo is betting that continued AI advances will require not only more powerful perceptual intelligence models, but “a continuous feedback loop” featuring new and increasingly better datasets and benchmarks. Feature image courtesy Y Combinator via LinkedIn.