Ex-Anthropic researchers launch Mirendil, target $175M at $1B valuation for AI-powered scientific discovery

A new wave of AI startups is taking shape, and Mirendil is stepping into it with serious ambition—and serious backing. Founded by former Anthropic researchers, the San Francisco-based startup is already in talks to raise $175 million at a $1 billion valuation, signaling strong investor appetite for AI applied to scientific discovery.

Mirendil is led by Behnam Neyshabur, who previously headed Anthropic’s scientific AI reasoning team, and Harsh Mehta, a former senior research scientist at the company. The two left Anthropic in December 2025, joining a growing list of researchers breaking away from major labs to build more focused AI ventures. Their founding team includes Shayan Salehian, formerly of xAI, and Tara Rezaei, a former OpenAI intern, bringing together experience from some of the industry’s most influential AI organizations.

From Anthropic to Neo-Labs: Mirendil Targets $1B Valuation to Reinvent Scientific Research with AI

The company is focused on applying AI to biology and materials science, two areas where progress often moves slowly. By building models that generate hypotheses and simulate outcomes, Mirendil aims to shorten the path from idea to discovery. If it works, the payoff could be significant—ranging from new drug candidates to advanced materials that reshape industries.

Funding talks are ongoing, with Andreessen Horowitz and Kleiner Perkins leading the round, according to reports citing The Information. The terms have not been finalized. The early valuation reflects growing investor confidence that domain-specific AI startups can deliver meaningful breakthroughs, especially in fields that have long relied on trial-and-error experimentation.

“Former Anthropic researchers launch Mirendil, aiming for a $1 billion valuation in AI-driven science,” The Information reported.

Mirendil fits into what many are calling the “neo-labs” movement, a shift where researchers leave large AI organizations to build startups focused on specific use cases. Instead of building general-purpose models, these companies are targeting narrower problems, expecting depth to outperform breadth. That shift is starting to reshape how AI innovation is distributed across the industry.

The timing is no coincidence. Large AI labs have drawn intense attention in recent years, along with growing pressure around safety, governance, and long-term direction. For some researchers, startups offer a faster path to execution and a clearer focus on real-world applications. Mirendil’s leadership team appears to be betting that specialized models trained on scientific data can unlock insights that broader systems struggle to reach.

If the funding round closes, Mirendil will have the resources to scale quickly, hiring researchers and expanding compute capacity to train its models. The next 12 to 18 months will be critical. Early partnerships, research outputs, and model performance will determine whether the company can stand out in a space that already includes heavyweights like Google DeepMind.

The bigger story may be what Mirendil represents. The rise of these neo-labs suggests a shift away from centralized AI development toward a more fragmented ecosystem, where smaller teams pursue targeted breakthroughs. That fragmentation could accelerate progress across industries, though it may also intensify competition for talent among the largest labs.

For now, Mirendil is positioning itself at the intersection of AI and scientific research, backed by a team that has already worked at the frontier. The market is watching closely. If the company delivers on its promise, it could help define how AI contributes to scientific discovery in the years ahead.