Problem

Cellular Intelligence, formerly Somite AI, targets a core bottleneck in cell therapy: a living-cell product’s identity, purity, potency, and viability depend on the signals, timing, and conditions used to make it. Conventional experimentation and manufacturing are costly, slow, difficult to scale, and vulnerable to process variability; today’s two-dimensional processes can scale out but not efficiently scale up, increasing development cost and failure risk. The company claims its data-generation approach can reduce the cost of studying cell-state transitions by roughly 1,000x, potentially compressing development timelines and expanding patient access.

The clearest near-term use case is designing and scaling reliable manufacturing protocols for stem-cell-derived therapies. Cellular Intelligence’s initial clinical demonstration engine is STEM-PD, an allogeneic Parkinson’s cell-therapy program acquired from Novo Nordisk, where the company can use its models to improve differentiation, comparability, scale-up, and product consistency while addressing a disease with substantial unmet need.

Product / Service

The company is building an integrated wet-lab and computational platform centered on a universal virtual cell-signaling foundation model. Its proprietary miniaturized capsule technology applies standardized combinations of signals to many starting cell states and reads the resulting cell states with RNA sequencing, generating millions of time-resolved perturbation conditions at much greater efficiency than conventional methods. Those data train predictive models that estimate how cells will respond to signals, schedules, drug combinations, and manufacturing conditions before each experiment is run.

The disclosed delivery model is full-stack rather than a conventional software-only product: Cellular Intelligence generates its own biological data, trains the model, and applies it to internal and potentially partnered therapeutic programs. The intended benefits are rational design of differentiation protocols, context-specific drug-response prediction, disease modeling, and more controllable manufacturing. The company’s model-to-medicine strategy also creates a data flywheel in which clinical and manufacturing data from owned therapies improve the platform and provide real-world validation.

Market

Cellular Intelligence competes in AI-native TechBio, virtual-cell modeling, regenerative medicine, and cell-therapy development. Its closest conceptual peers are companies and initiatives building virtual-cell or foundation models, including Noetik’s cancer-focused virtual-cell work and broader efforts associated with the Chan Zuckerberg Initiative, Google DeepMind, and the Arc Virtual Cell Atlas. It also competes indirectly with conventional cell-therapy developers and AI drug-discovery platforms, but the available evidence does not establish a definitive like-for-like commercial competitor.

The company has substantial financing and clinical traction rather than being only a research-stage startup. It closed a more than $47 million Series A led by Khosla Ventures in May 2025, received a strategic investment from AMD Ventures, and reported more than $62 million raised from Khosla, the Chan Zuckerberg Initiative, SciFi VC, AMD Ventures, Fusion Fund, and others. In May 2026 it secured global rights to Novo Nordisk’s STEM-PD program; the therapy has an FDA-cleared IND for Phase 2, Fast Track Designation, and underlying Phase 1/2 dosing. Novo Nordisk also became an equity investor. The reviewed materials do not disclose revenue or commercial sales, so the company is best described as clinical-stage with revenue status undisclosed, not as demonstrably revenue-generating.