Solutions for cloud-hosted GPU farms for an interactive workshop

Follow the full discussion on Reddit.
I'll be delivering an interactive workshop on LLM fine-tuning and deployment. As part of the workshop, I'd love for the attendees to try some hands-on experiments running scripts and notebooks. I won't be able to provide the physical hardware itself, so I plan to lease GPUs from the cloud instead. I am wondering if there are any ready-made solutions available for this sort of use case. Ideally one where I can deploy a certain number of instances, perhaps that even scale with demand, and grant individual user access into isolated docker containers. Perhaps this is asking too much, but maybe there's something out there which gets me most of the way there. The backup would be building this system myself on a major cloud provider, which would take some time. Thanks!