In the rapidly evolving world of artificial intelligence, a new frontier is emerging. Mirendil, a San Francisco-based AI company founded in early 2026, is at the forefront of this revolution. The company is developing autonomous AI systems designed to accelerate scientific research across fields such as biology and materials science. This innovative approach promises to democratize access to cutting-edge AI research, enabling more scientists and organizations to tackle complex problems.
The vision behind Mirendil is to create AI systems that can generate hypotheses, run experiments, evaluate results, and iterate on findings autonomously. This self-accelerating AI research and development (R&D) is seen as the most direct path to delivering on the broader promise of AI. By automating the research loop, Mirendil aims to fundamentally change the rate of progress across all domains.
Mirendil’s Unique Approach to AI Research
Most AI labs are research organizations that happen to use AI. Mirendil, however, is the inverse—a frontier lab rebuilt from scratch around AI. The company trains frontier models that are exceptional at AI R&D and redesigns the entire loop around them to make it faster, more capable, and more autonomous. This approach creates a virtuous cycle where better models do better research, and better research produces better models.
The name Mirendil, meaning ‘friend of precious things, the hidden and the undiscovered,’ reflects the company’s mission to uncover new scientific breakthroughs. The team behind Mirendil includes some of the most experienced researchers and engineers in the field, with backgrounds from leading institutions such as Anthropic, xAI, Google DeepMind, and OpenAI.
The Founding Team and Their Vision
Mirendil was founded by Behnam Neyshabur and Harsh Mehta, who met seven years ago at Google through a cold email. Behnam is the co-inventor of the SAM optimizer and has done foundational work on why deep learning generalizes. Harsh initiated and led the effort for automated AI R&D at Anthropic. Together, they have co-led efforts to build models that can reason through hard math and code, such as Mirerva and Gemini.
The founding team also includes Shayan Salehian, an early engineer at xAI, and Tara Rezaei, a 23-year-old MIT graduate and early student researcher at OpenAI. This diverse and experienced team is united by a passion for science and a drive to build technologies that move it faster.
Mirendil’s Impact on Scientific Research
Mirendil’s goal is to put frontier AI R&D in everyone’s hands, allowing the people closest to a problem to stay focused on solving it. This democratization of AI research can accelerate scientific progress across various fields, from drug discovery to robotics. By enabling more scientists and organizations to access and control their own AI infrastructure, Mirendil aims to pull concentration of power away from a few labs.
The company is still small, with a narrow focus on hiring top technical talent in San Francisco. Mirendil’s approach is to compound, not grow, ensuring that the systems they build deliver real value to scientists as soon as possible. The potential impact of Mirendil’s work is immense, with the possibility of accelerating scientific discovery and solving some of humanity’s most pressing problems.
