Michael Polansky is training an AI model on skin that’s still alive

The surface story is a celebrity-adjacent startup announcement, but the underlying engineering problem is far more interesting. Keeping human skin alive outside the body for weeks is not a trivial perfusion problem—it requires mimicking the dermal-epidermal junction, nutrient exchange, waste removal, and mechanical stress profiles, all while preventing necrosis and microbial overgrowth. If Polansky’s team has actually solved that, they’ve built a powerful assay platform. The hook to AI is likely in pattern recognition: applying ML to multimodal readouts (histology, gene expression, barrier function) to predict compound efficacy and toxicity earlier than current models allow. That’s a legitimate scientific frontier, though the press release style often overstates maturity. But the tech world should be skeptical of the "AI + biotech" combo when it comes from a tech investor rather than a bench scientist. Polansky’s background is in technology and capital, not dermatology. That doesn’t disqualify him—many successful biotech startups are led by operators who hire domain experts—but it shifts the burden of proof. The real validation will come from peer-reviewed data or a large pharma partnership, not from TechCrunch profiles. The claim of "weeks" of viability is also crucial: most ex vivo skin models last days. If true, that’s a breakthrough; if it's about maintenance of a specific skin type or partial viability, the gloss hides the nuance. There's also a broader theme here about the convergence of AI, tissue engineering, and personalized consumer products. Skincare is a high-margin market with notoriously weak evidence for many active ingredients. A platform that can rapidly test compounds on living human tissue—before spending years on animal or human trials—could reduce the cost of innovation and even enable personalized formulations based on a patient's own skin. That's a compelling vision, but it assumes the ex vivo model accurately recapitulates in vivo responses, which is still a major open question. Finally, the celebrity aspect (Lady Gaga’s partner) will inevitably distort coverage. That’s fine for fundraising optics, but it doesn't change the science. What matters is whether the company can reproduce results, share protocols, and eventually pass regulatory muster. The industry has seen too many "revolutionary" skin-tech startups that fade after a splashy launch. The source URL (https://techcrunch.com/2026/08/21/michael-polansky-is-training-an-ai-model-on-skin-thats-still-alive/) provides the initial details, but we need more depth—technical documentation, validation data, or independent replication—before we can judge if this is a genuine leap or a carefully pitched demo.