Mirendil is expanding its use of Google Cloud’s AI Hypercomputer to support the pre-training and post-training of advanced artificial intelligence models. The frontier AI research company will use a combination of Google Tensor Processing Units and NVIDIA accelerated computing infrastructure running on Google Cloud.
Mirendil plans to use the infrastructure for complex model-training workflows, including initial pre-training, post-training and large-scale reinforcement learning.
The company is developing AI systems designed to accelerate the research process itself, helping scientists and engineers design experiments, evaluate results and iterate more efficiently.
Access to both TPUs and NVIDIA GPUs will allow Mirendil to match individual workloads with the computing architecture best suited to each task.
Google Cloud worked with Mirendil on the design and deployment of the combined infrastructure across computing, storage, networking and control planes.
The companies also collaborated on a system that uses managed training clusters running through the Gemini Enterprise Agent Platform.
The system is intended to streamline the provisioning and management of Mirendil’s TPU and GPU environments, reducing the operational complexity associated with large-scale AI training.
Mirendil is already operating a cluster powered by Google Cloud TPU v5P chips. NVIDIA accelerated computing systems are expected to come online soon.
The deployment gives Mirendil access to flexible computing resources as it develops AI systems capable of improving and automating portions of the research cycle.
The company’s broader objective is to make frontier AI research capabilities available to a larger group of scientists and engineers.
Google Cloud said its AI Hypercomputer provides purpose-built infrastructure spanning accelerators, software, networking and storage for model training, inference and advanced research.
The Mirendil deployment reflects increasing demand among AI laboratories for infrastructure that can support multiple accelerator architectures rather than requiring every workload to run on a single type of chip.
KEY QUOTE:
“Progress in AI has been bounded by how fast humans can run the research loop, designing experiments, evaluating results and iterating. We’re building AI systems that can accelerate and improve that loop itself.”
“Expanding on Google Cloud gives us the scale and flexibility to push those systems further and put frontier AI research capabilities in the hands of many more scientists and engineers to run that loop faster and at a greater scale.”
Behnam Neyshabur, Co-Founder and CEO of Mirendil
