Cellular Intelligence Unveils with “AlphaFold Vision” for AI-Driven Cell Therapy

The year was 2021. Tech entrepreneur Micha Breakstone, PhD, had just sold the conversation intelligence platform, Chorus.ai, to ZoomInfo for $575 million and was eager for a new challenge. As an expert in natural language processing and automatic speech recognition, Breakstone recalled watching the fields commoditize and was not interested in another “incremental technical win.”

“I was brought up with the calling of Tikkun Olam,” said Breakstone in an interview with GEN Edge, as he described the Hebrew phrase, “repairing the world.” 

“Biology felt like the one place where I could truly live that calling,” he continued.

Breakstone was prompted to reconnect with a long-time friend of 27 years, Allon Klein, PhD, an esteemed stem-cell biologist and now professor of systems biology at Harvard Medical School (HMS). Breakstone “knew nothing about biology at the time,” but was intent on learning the science and culture of biotech, even as Klein remained skeptical and hesitant to mix business with a friendship that began at age 19.

Klein warmed up to the idea of partnering on a new enterprise after Breakstone “came back twice more,” and facilitated introductions to Harvard developmental biology colleagues and National Academy of Science members, Olivier Pourquié, PhD, and Cliff Tabin, PhD.

What slowly came together was a mission to pursue what some see as a holy grail of regenerative medicine: the ability to generate any human cell type to replace diseased or damaged tissue.

More than somites

Remarkably, only 20 fundamental molecular signaling pathways give rise to thousands of cell states, resulting in an unfathomably large search space when researchers attempt to engineer a particular cell-type. Less than one percent of known human cell types can be reliably produced for downstream applications in cell therapy.

“Every cell that we discover or optimize opens a slew of potential applications,” said Breakstone. “One could spend a decade and tens of millions of dollars on painstaking trial-and-error to differentiate a new cell type, or solve this problem in one fell swoop, much like AlphaFold for the protein folding challenge.”

In January 2024, Breakstone and the Harvard team launched Somite AI. The company name was originally inspired by the goal of applying somites, the developmental structures at the origin of the musculoskeletal system, in cell replacement therapy for conditions such as Duchenne muscular dystrophy (DMD), a severe, progressive genetic disorder causing muscle degeneration.

Somite AI has now re-branded under the new name Cellular Intelligence, with a mission to transition from empirical trial-and-error to AI-guided rational stem cell engineering.

The team is building what they describe as the first “universal virtual cell signaling model,” capable of learning the “grammar” underlying how sequences of signaling cues lead to cell differentiation. This “AlphaFold for developmental biology” would theoretically guide the production of any cell type on demand, dramatically broadening the landscape of achievable cell therapies.

Among the company’s therapeutic areas of focus include developing insulin-producing beta cells as a potential curative treatment for type 1 diabetes, satellite cells for neuromuscular disorders, and brown adipocytes for metabolic disorders and longevity.

Following a $47 million Series A in May 2025, Cellular Intelligence has raised a total of $62 million from Khosla Ventures, CZI, AMD Ventures, Astellas, and more. Additional scientific co-founders include Jay Shendure, MD, PhD, Howard Hughes Medical Institute (HHMI) investigator and professor of genome sciences at the University of Washington, and Jonathan Rosenfeld, PhD, head of fundamental AI at Massachusetts Institute of Technology (MIT), who serves as chief technology officer at Cellular Intelligence.

Big data from a capsule 

Unlike the protein folding problem, where AlphaFold thrived by learning from the extensive Protein Data Bank (PDB), datasets that capture the context-dependent and temporal process of stem cell differentiation remain sparse.

Pourquié told GEN Edge that Cellular Intelligence is making progress toward large-scale, causal, and sequential perturbation data across many developmental states to allow models to predict state transitions.

The company’s platform leverages a semi-permeable capsule technology, recently published in Science by Klein and his colleagues, which selectively retains cells and large analytes while being freely accessible to media, enzymes and reagents. The method enables high-throughput assays combining live-cell culture with genome-wide readouts. Millions of time-varying signal combinations are tested on human stem cell differentiation in parallel, providing 1,000 times higher efficiency than traditional methods.

“It’s like we’ve gone from exploring one coastline at a time to now having satellites capture the shape of the entire globe,” Klein told GEN Edge. 

Notably, the company’s emphasis on cell signaling is distinct from other virtual cell players, including Biohub (formerly Chan Zuckerberg Initiative (CZI)), Xaira Therapeutics, and Arc Institute, which have focused on understanding internal genetic circuitry. These modeling efforts are guided by large scale observational or perturbation data (drug or genetic screens) describing gene expression in a limited number of cell types, often artificial cancer cell lines.

In contrast, Cellular Intelligence is building an AI platform that allows the generation of thousands of distinct biological contexts, described by intermediate developmental states, using human stem cells.

Although a full virtual cell signaling model is still underway, Breakstone asserts that early data already demonstrates that targeted pathway perturbations can steer cell state. “It’s a small step for machine learning,” he notes, “but a big step for developmental biology.”

Breakstone’s friendship with Klein remains strong. “Allon was the smartest person I knew then, and probably still is,” he said. “I always had this dream that we would build something together one day.”