What if we could train AI to understand how stem cells become any cell type in the human body? In this episode of the In Vivo Podcast, host David Wild sits down with Micha Breakstone (CEO & Co-founder) and Samantha Dale Strasser (VP of Strategy) from Somite.AI to explore how their company is using foundation models to revolutionize cell therapy development. Somite has pioneered a breakthrough approach that generates cell differentiation data at 1000x the efficiency of traditional methods using proprietary hydrogel capsule technology. By capturing millions of trajectories showing how cells respond to signals over time, they're building DeltaStem—a foundation model that could do for developmental biology what AlphaFold did for protein structure prediction. Topics covered: - How bringing cells to signals (rather than signals to cells) unlocks exponential scale - Why wet lab innovation is just as critical as AI models - Manufacturing optimization: improving purity, reducing variability, and cutting costs - From beta cells for diabetes to brown fat for metabolic disease—the therapeutic pipeline - Why even AI experts underestimate what's coming in the next decade - Lessons from building biotech companies from academic concepts to commercial ventures Whether you're in pharma, biotech, AI or just fascinated by the intersection of technology and human biology, this conversation offers a grounded look at how foundation models are moving from hype to real therapeutic impact.