Mirendil is Scaling Self-Accelerating AI with Google

Mirendil, democratizing frontier AI R&D to accelerate science and technology, led by Behnam Neyshabur, Harsh Mehta, Shayan Salehian, and Tara Rezaei, announced a multi-year partnership with Google Cloud to power the next generation of self-accelerating AI. Mirendil will scale its training, inference, and AI research workloads on Google Cloud’s AI Hypercomputer, drawing on both Google Cloud TPUs and full-stack NVIDIA AI infrastructure.

Removing the Bottleneck in AI Progress.

AI progress has been constrained by the speed of the research loop: designing experiments, evaluating results, and determining what to test next. Mirendil is developing self-accelerating AI designed to eventually automate much of the work performed by frontier AI labs, continuously improving the research process and enabling faster, more autonomous AI development.

The long-term goal is to broaden access to frontier AI research and development, allowing teams in medicine, biology, materials science, and other fields to build specialized AI systems without requiring the infrastructure of the world’s largest AI labs.

Scaling the Research Loop.

Building self-improving AI systems requires increasingly complex workloads across large compute environments, spanning pre-training, post-training, large-scale reinforcement learning, and parallel research experiments. Mirendil’s systems are designed to run across different accelerators, optimizing the software stack above the chip layer and matching workloads with the most effective available compute.

Google Cloud’s AI Hypercomputer, including managed training clusters co-designed across compute, storage, networking and control planes, provides access to TPUs and NVIDIA’s accelerated computing platforms at frontier scale. This enables larger and more sophisticated research loops while reducing the operational complexity of running reliable frontier-scale experiments for both researchers and AI agents.

Amin Vahdat, SVP and Chief Technologist for AI and Infrastructure at Google, said AI advancement is increasingly dependent on orchestrating entire systems rather than improving chip-level performance alone. He noted that Google Cloud’s AI Hypercomputer combines co-designed hardware, networking, and software across a broad portfolio of TPUs and GPUs, enabling Mirendil to run complex training workloads more efficiently and accelerate research loops aimed at advancing scientific discovery.

The partnership strengthens Mirendil’s relationship with Google and expands its compute infrastructure across leading providers. The collaboration supports the broader goal of making frontier AI research more accessible and enabling teams to tackle increasingly complex scientific challenges.

𝐑𝐞𝐦𝐨𝐭𝐞 𝐇𝐢𝐫𝐞 𝐖𝐢𝐭𝐡 𝐔𝐬: 𝐌𝐞𝐧𝐥𝐨 𝐓𝐚𝐥𝐞𝐧𝐭 helps technology companies hire exceptional remote talent from India. Build your team with carefully curated engineers, product leaders, designers, GTM professionals, and more. Learn More At: https://www.menlotimes.com/menlo-talent