In this blogpost, we’ll spotlight a new set of new Change Management features we’ve added to our flagship product,ScienceOps.
ScienceOps is a platform for deploying and managing predictive models in production applications. ScienceOps lets data scientists deploy R and python models into web and mobile applications without tossing their work “over the fence” to the dev team.
What's Change Management?
While ScienceOps makes it much easier to deploy predictive models into production, data science teams don’t always want all users to be able to push code directly into production. Change Management is a set of features that helps determine how models go from development to production.
Read-only API Analyst Manager Admin Query a model x x x x Deploy models x x x Run tests x x x View shared models x x x Submit models for approval x x x Approve models for production x x Create/delete users x How's it work?Anthony Analyst has just written a Python product recommender model and is eager to get his code into production. He deploys his model to development, to his user account endpoint /anthony/models/ProductRecommender and submits his model to his manager for approval.
Manager Madeline reviews Anthony’s model and has a few ideas for improvement. Manager Madeline rejects Anthony’s model and asks him to try adding a couple new variables, like trending purchases.
Anthony makes the changes (he happens to be writing his model his favorite Python IDE,Rodeo. Anthony deploys the new version of his code and resubmits it for approval, though his manager, Madeline is out for the week (Madrid? Madagascar?! Madison??)
Since his manager is out of town, Anthony reaches out to Adam Administrator, who has the authority to approve any model into production. Adam checks out Anthony’s revised model, which is currently deployed to Anthony’s development environment, and approves it to be deployed into the production environment.
Technically speaking, that means that Anthony’s model becomes available at the endpoint production/models/productrecommender once it is has been approved. Anthony’s Python product recommender model is now being called from the company’s website and is generating recommendations in real time!
Model Read-Only API KeysWe’ve also added read-only API keys to ScienceOps for increased security. This allows you to use a unique API keys to grant apps read access to models deployed to ScienceOps.
Wrapping Up
We hope you like these new Change Management features as much as we do!
If your company is interested in learning more about deploying data science models usingScienceOps, click to schedule a demo below!