Hi Rantsho,

Building your performance prediction application is no different than what I’ve discussed in this article.

Draft up a few simple sketches of your envisioned app, code your model inside a Jupyter Notebook, add endpoints to your Jupyter cells where needed, and build a simple web application that calls your endpoints and displays the necessary information. If you want to test this app among many users, you can put it inside a Docker container and ship it to a server.

I would also decide if you need to call the model each time a prediction is made, or if it would make more sense to calculate predictions in batch and have your web app read the saved prediction file.

In any case, the main steps would be the same as discussed in this article. Let me know if you still have questions. Importantly, don’t feel you need a detailed plan before starting; just start building and be willing to make mistakes. It’s a far superior way to learn than to follow a recipe.

Keep me posted.