I’ve been reviewing a cloud automation platform that includes a cool way to work with Jupyter Notebook. I thought it could interest you! It’s called MMCloud from MemVerge.
MMCloud allows you to run a Jupyter kernel (or anything else) on a cloud instance and not worry about the infrastructure. If the instance stops (e.g. spot instances), if you encounter an out of memory issue (like loading a large dataset) or if you use up too much CPU, then the service saves the state of the instance and you can migrate automagically to another instance. The migration policy is fully customizable.
If you are like me and made an ever loaded too much data or relied on spot instances to do your exploration and had your kernel dying on you causing you to lose some work , then you will love this. See this demo:
MMCloud is deployed in your cloud VPC as a single tenant, so data privacy is built in (no sharing). They can also run the service and have a free tier. They told me that they will soon release a built-in Jupyterhub integration which should make things even simpler
If there is interest, I am happy to make a live demo. Posting here a link to a form with the details: https://forms.gle/NgJKJqG3BkB243o28
I am happy to answer any questions.
Disclaimer, while I work at Quansight and I am consulting with MemVerge, I am very enthusiastic about their solution! I personally endorse their solution after having done my own evaluation and tests. I found it very straightforward to setup and it delivered on the promises.