Understanding the value of model serving patterns

Using patterns for ML model serving make us more productive in bringing our model to clients. If we do not follow any patterns, then we may struggle to find the right tool and strategy needed to serve the model for a particular problem.

Figure 2.1 – Alice needs to perform trial and error with multiple tools to find the right one

Let’s consider the situation of Alice in Figure 2.1. Alice has a problem that involves making a data-driven decision. She needs to create a model to solve the problem and deploy the model using a serving tool. She has thousands of tools on offer. She needs to study all these solutions and find the best solution. There is another challenge in the approach of selecting the right tool. Alice is at risk of making a bad choice of tool, as she is solving an optimization problem manually and can be stuck at local maxima. This is always an impediment to productivity, as it...