dimzachar/xGoals-mlops ? reverse-engineered prompt

Build me an end to end project for football expected goals, where I can pull in shot event data, train a simple xG model, and then deploy it in a way that feels production ready.

I want the repo to include a notebook for exploring the data and training the model, plus a proper pipeline that handles data loading, preprocessing, training, evaluation, and saving the model. It should work with the JSON event data from the football dataset in the repo config, and be set up so I can rerun everything cleanly if the data link changes.

Please also wire up the MLOps side so it can be run in Docker, tested automatically, and deployed with AWS style infrastructure support, including a serverless function for predictions and basic CI CD for checks. If it helps, use the existing config files and scripts already in the project, and look up current docs online if you need to. I’d like the whole thing to be easy to reproduce, easy to test, and clear enough that I can extend it later.