uber/causalml — reverse-engineered prompt

Build me a Python library for causal inference and uplift modeling that helps data scientists estimate how much a treatment changes an outcome for each user, using features, treatment assignment, and outcome data. I want a clean, consistent API so someone can train models on randomized experiments or observational datasets, then get conditional average treatment effect estimates, uplift style scores, and simple recommendations for who to target or which treatment to choose.

Please include a few standard methods people would expect for this kind of work, plus solid documentation, a quickstart, and example notebooks that show realistic use cases like ad campaign targeting and personalized customer engagement. It should feel like a real open source package, with tests, installable packaging, and docs that are easy to browse. If some methods are still experimental, label them clearly so the more stable parts are easy to trust. Use Python throughout, and look up current docs online if you need to.