LFS302 - Building Scientific Workbench Environments
1 min read
Claims (Hypotheses)
Scientists need a scalable workbench environment. I’m actually not sure I agree with this.
Emphasis on biology
Deploying custom algorithms (analysis algorithms, not processing algos)
Challenges
Easy and efficient access to data
Accessing third party data securely
Allowing scientists to experiment with their data but selectively deploy working algorithms and workflows
Catering to scientists who are not experts in coding or ML
Carrying Jupyter notebooks into production deployments (hmmm…)
Blocker and Challenges
What is our biggest challenge? Providing scientists a good prototyping environment that we can transition into the production processing environment. Workspaces might be a good solution. Multi-tenant Jupyter could be another.
Workspaces: lowest common denominator that works for any needs, provided the scientist knows how to set up their own dev environment.
Jupyter: Easier to use but limited. Scientists will need to use a separate environment to package their code into Docker.
