Model monitoring without modifying existing code

[iamavi9]

Hi Team.

We have an ML model deployed as a REST API endpoint in GCP CloudRun service. We want to monitor the endpoint and effectively validate drift detection. May I know if this is possible without modifying my code. I am looking for a plug-n-play option if available. Also, let me know if EvidentlyAI has any enterprise offering that covers my requirement.

[elenasamuylova]

Hi @iamavi9,

Since Evidently is open-source, it requires you to manage the integration on your own. You can either:

• Run scheduled monitoring jobs over the model monitoring logs (e.g., predictions stored in the data warehouse)
• Host the Evidently collector service on your own and send predictions from your ML service

You can also run ad hoc checks e.g. by performing analysis (generating Evidently Reports) in Jupyter notebook.

We are working on the Evidently Cloud to simplify the integrations and provide additional enterprise features. It is not available publicly, but you can sign up for the waitlist here https://www.evidentlyai.com/cloud-waitlist.