How do I get beyond just fitting models in Python?

Try taking one question all the way through instead of starting with a model. Find a dataset, decide what you want to learn from it, inspect missing values and odd records, then explore the data before choosing a method.

For a prediction task, decide how to split the data before tuning anything, compare against a simple baseline, and look at where your model makes mistakes. Finish by explaining what the results do and don't tell you. That whole process is useful practice, even if the final model is pretty simple.

Take a look at data engineering too. Models need data that can be collected, cleaned, and refreshed reliably, and that's a different skill from calling fit in a notebook.

You could start by moving data into a database and building repeatable transformations. From there, dbt and Airflow are tools worth exploring. Pick a cloud platform later if you want to practice deployment, and look at Terraform when you're ready to manage infrastructure.

I'd spend some time on data analysis before adding more modeling libraries. Practice answering questions with summaries and plots, checking assumptions, and explaining uncertainty. A data mining course could also give you a wider view of methods for finding patterns, though it doesn't have to be your next stop.

If you want structure, try a data science learning roadmap and use it as a checklist, not a rulebook. Pick the areas you haven't practiced yet, then build a small project around each one so you're not just collecting courses.