TileBio Information

Most pathology AI is only as good as the labels it was trained on. TileBio's learns straight from the tissue - no labels to limit what it can uncover. Our foundation model learns the language of tissue directly from millions of unlabelled whole-slide images, discovering morphology, phenotypes, and tissue-specific context no annotation scheme has ever defined. The result is a model that can both recognise categories it was trained on, and - more importantly - flags the morphology and insights nobody thought to look for. For toxicology and preclinical safety teams, this means treatment-related findings are caught consistently across every study and site, including the subtle ones a manual scoring scheme is likely to miss. For discovery and translational teams, it means a full phenotype map of a cohort, surfacing biomarker candidates a predefined scheme would never have captured. For healthcare systems, it means applying the same unsupervised phenotyping at scale: resulting in faster, more consistent patient stratification and better use of specialists' time. TileBio is a University of Glasgow spin-out, built by a team spanning computer science, diagnostic histopathology, cancer surgery, and molecular medicine. We're backed by Twin Path Ventures, Maven Capital, Scottish Enterprise, and GU Holdings, and work alongside NHS Greater Glasgow & Clyde and Cancer Research UK. Curious what the model would find in your own slides? Start a pilot or propose a collaboration at tilebio.com.