mlops-project-requirement

[DavidSolan0]

This project aims to create a draft of the MLOps considerations for a company that has ~20 to 50 ML models running inference in real time for upwards of 100.000 users monthly.

Assumptions

1. All infrastructure is on AWS.
2. Assume you have an unlimited budget and can use any AWS service.
3. Assume each model runs on a standalone Kubernetes pod or EC2 instance for inference. Each of them is accessed through a REST API built on FastAPI.
4. All the Data Engineering has been done to get clean data for any given model to a Redshift DataWarehouse.

Project setup and goals

• Create UML diagrams #6

• Cover Inference Scalability #2

• Create ML CI/CD #3

• Drifting Evaluation #4

• Build CT #5