Mirendil, a frontier AI lab focused on accelerating AI research and development, has selected Google Cloud's AI Hypercomputer infrastructure. The partnership gives Mirendil access to a mix of Google TPU v5P chips and NVIDIA GPU accelerators to support large-scale model pre-training and post-training workflows, including reinforcement learning. Google Cloud collaborated with Mirendil on end-to-end design and deployment using managed training clusters running in the Gemini Enterprise Agent Platform.