Webinar
Your DS team just let you know that instead of one model serving predictions for all your customers, from here on out each customer will have their own separate model. Now your MLOps team needs to support multiple models spanning, training, validation, serving, and monitoring - your pipeline just became x10 more complex than it was just a day ago, and scaling your architecture means more than just adding a replica.
This session will cover architectural considerations for multi-tenancy in ML, best practices in traditional software engineering that can be copy/pasted over to MLOps, as well as new considerations unique to ML
- Why and when to use ML multi-tenancy
- Architecture pros and cons
- Observability and monitoring at high scale
- Security and compliance considerations
