Kubeflow, MLflow and Argo on On‑Prem HPC for Dementia Research at Imperial College UKDRI

Imperial’s UK Dementia Research Institute runs complex, regulated AI pipelines on sensitive clinical and home‑monitoring data, making “just use a managed cloud ML platform” a non‑starter. This talk presents a reference architecture for running Kubeflow, MLflow and Argo Workflows on on‑prem HPC to support dementia research, from raw sensor and imaging data through to trained models and reproducible analyses. It walks through how researchers submit experiments that span Slurm‑backed GPU/CPU nodes and Kubernetes‑hosted services, while keeping data within institutional control and traceable for ethics and governance review. Attendees will see practical patterns for experiment tracking, workflow orchestration, and model promotion that fit the realities of public‑sector research IT: constrained budgets, strict information‑governance controls, and long‑lived cohorts of vulnerable participants.

Decoupling from the Hyperscalers: A Kubernetes-Based Sovereign AI & Compute Platform

As AI and compute workloads grow, organizations are seeking greater control, security, and independence from hyperscaler cloud providers. This talk explores how we built a Sovereign AI & Compute Platform entirely on Kubernetes, designed to operate across diverse environments without vendor lock-in. We’ll dive into the architectural principles, open source components, and operational patterns that enable scalable, cloud-agnostic compute - from infrastructure orchestration to AI workload management. Attendees will learn how Kubernetes can serve as the foundation for a sovereign, portable, and resilient cloud-native ecosystem.
Built on this foundation, stack8s integrates with 15+ cloud providers and 700 data centers, allowing customers to keep their data where it is. We bring compute closer to compliant data through encryption, low-latency connectivity, and a true hybrid/multi-cloud experience. No more infrastructure headaches. No more vendor lock-in. Own your compute end to end.

As AI and compute workloads grow, organizations are seeking greater control, security, and independence from hyperscaler cloud providers. This talk explores how we built a Sovereign AI & Compute Platform entirely on Kubernetes,

Realising Compute as a Commodity: Releasing Innovation Beyond Hyperscalers

For too long, infrastructure decisions have dictated innovation velocity. Organisations build around cloud providers instead of building around outcomes. At stack8s, we believe compute should be treated like electricity: available, interchangeable, and priced competitively. The value should sit in the control plane, the architecture, and the workloads, not in the underlying provider. By abstracting infrastructure through Kubernetes and a unified control layer, stack8s enables organisations to consume GPU and CPU resources across multiple clouds and sovereign environments as interchangeable capacity. This shifts power away from hyperscaler dependency and back to the organisation.

In practice, this means researchers, enterprises, and AI teams can select the right compute for the job based on performance, sovereignty, compliance, or cost, without redesigning their architecture each time. stack8s orchestrates distributed worker nodes across regions and providers into a single operational fabric, enabling multi-cloud AI workloads, regulated data handling, and cost arbitrage of GPU markets. When compute becomes a commodity, innovation accelerates. Teams focus on models, data, and outcomes, while infrastructure becomes fluid, competitive, and sovereign by design.

Bringing Compute to FAIR Data: A Modern Platform for Al-Driven Research

UK Dementia Research Institute FAIR Workshop