Case Study: Simulacra AI achieves 90% faster model compilation with Nebius infrastructure

Quantum Chemistry for drug and material discovery

Simulacra AI was faced with the challenge of building a scalable foundational wave-function model for molecular systems. Their goal was to generate high-accuracy datasets for drug and material discovery pipelines but they needed to overcome significant scalability and efficiency hurdles. They chose to partner with vendor Nebius to leverage its infrastructure for this task.

By implementing the Nebius infrastructure, which features a fleet of NVIDIA Tensor Core GPUs including H100 and H200 models, Simulacra AI gained a highly efficient solution. This resulted in their largest models, with over 100 million parameters, now compiling for pre-training in just 10-20 minutes, a 90% improvement that previously took over two hours. With Nebius, Simulacra AI can now deliver next-generation molecular data, enabling companies to refine their in silico pipelines without needing vast internal infrastructure.

Simulacra AI