How they use Nebius AI Cloud

TrialHub uses two NVIDIA L40S GPUs with Intel Ice Lake on Nebius AI Cloud to generate vector embeddings from medical and clinical-trial information. The customer story reports 100% GPU utilization. With Nebius expert support, it launched a production-ready, 250-million-vector database in days. Its MLOps-optimized stack integrates embedding workflows with a growing medical database. Raw inputs are stored in MongoDB and the embedding database in Zilliz. The customer story describes RAG-optimized LLMs and semantic search across more than 80,000 trusted medical sources, supporting research answers in its clinical-trial planning platform. Medical experts select sources for each use case, including medical guidelines, national reimbursement registries and peer-reviewed research articles. Semantic search retrieves disease information across countries to inform epidemiology and patient journey analysis.

TrialHub says traditional feasibility assessments took weeks of research through scattered medical guidelines and regulatory documents. Teams spent months synthesizing text-heavy sources, consulting experts and making judgments based on proxy data. Standard of Care assessments for just 10 countries could occupy the whole team for a week.

Nebius helped TrialHub meet tight deployment deadlines without in-house DevOps. The customer story reports domain-specific answers in seconds, 50% less time spent on feasibility assessments and three times faster patient recruitment. Nebius’s deployment post, shared by TrialHub’s company account, reports that trial delays and amendments were reduced by half. Nebius’s healthcare page reports 20 times faster research. TrialHub’s company post says Standard of Care research that took 20 hours now happens in minutes.