A complete guide to building a Docker Image serving a Machine learning system in Production

Building a Docker image is generally considered trivial compared to developing other components of a ML system like data pipeline, model training, serving infra, etc. But an inefficient, bulky docker image can greatly reduce performance and can even bring down the serving infra. This blog aims to focus on building an ideal Docker image and not on its concept or benefits. Most of the time a ML system will be based on Python, so it critical building any Python-based Docker image efficiently. Let us go through them.