Large language models can read text and images, but they do not know how a tablet dissolves, how a flavor lingers or how a material wears down. Apoha, a UK startup emerging from stealth on June 3 with $36 million in Series A funding, is trying to fill that sensory gap.
Its VIBE system suspends material samples in liquid, applies controlled physical stress and records the resulting vibration patterns. In minutes, Apoha says one sample can yield more than 1,000 behavioral measurements that once took days to collect, then convert those signals into AI-ready behavioral embeddings.
The company calls the approach "liquid intelligence." COO Anshika Srivastava says AI has not yet learned to taste, smell or feel matter, and Apoha wants to model that layer for drug dissolution, flavor performance and material wear. Early use cases include screening drug failure risk before expensive clinical trials, finding plant-protein substitutes and ranking antibody candidates with reported accuracy above 90%.
The round was led by Singular, with Draper Associates, Redalpine, Seedcamp, Wilbe, Nucleus and Innovate UK also involved. Apoha is still a small team of about 25, led by mechanical engineer and former Oxford postdoc Shamit Shrivastava, but it is betting on data that the open internet never gave language models.
Sources：CocoLoop、Apoha emerges from stealth with $36 million Series A round (Fortune)