Deepcell and Gilead Sciences entered a partnership to codevelop a custom AI foundation model for Chinese hamster ovary (CHO) cell line development using cell morphology data. Under the collaboration, Deepcell will image Gilead’s CHO lines on its REM-I high-dimensional cell morphology platform and train the model using self-supervised learning. The company said the approach aims to deliver earlier readouts on cell line quality—attributes tied to titer, stability, aggregation, and metabolic profiles—without requiring manual labeling. Deepcell described the training concept as leveraging its Human Foundation Model work and adapting it to the specific needs of cell line development teams. Deepcell and Gilead plan to evaluate the initial model on held-out Gilead cell lines and publish an open-weight version of the model, according to the press announcement. The companies did not disclose IP ownership details or financial terms. For bioprocessing and platform players, this is a signal that morphology-driven foundation models are moving from general analytics toward application-specific co-development with major biologics manufacturers.