Running Secure Life Science Research at Scale using Hybrid GPU HPC and Kubernetes 🧬

AI is transforming life sciences research at institutions such as Imperial College London, but there is a hard infrastructure reality: most research institutes do not have enough on-prem HPC or GPU capacity to meet the growing demand from AI workloads. Researchers need secure access to serious compute, but sensitive data, governance, reproducibility, and sovereignty cannot be an afterthought. 🚀

This session shows how stack8s uses Kubernetes as a global control plane for sovereign AI research. Institutions can start with their own on-premises Kubernetes environment, then securely burst workloads into more than 30 Neo clouds without requiring researchers to manage individual cloud accounts. The result is a governed AI platform where data, policy, identity, and TRE controls remain under institutional control, while GPU and HPC workloads can scale globally on demand. 🌍

Drawing on real-world examples from UKDRI teams, including Imperial College London, working on Dementia, Parkinsons and neurogenomics, the talk will share practical patterns for building Kubernetes-enabled Trusted Research Environments, isolated research enclaves, notebooks, pipelines, private models, shared storage, GPU scheduling, and observability.