Experimentation & Causal Inference jobs

Someone has to say whether the new model actually made things better, and these roles own the machinery for answering: A/B platforms, switchback tests, and causal inference for the cases where you can't randomize. LLM features raised the difficulty — treatment effects on open-ended conversations resist tidy metrics, which is precisely why companies hire for this.

The toolkit covers CUPED and variance reduction, interference-aware designs for marketplaces, and libraries like EconML or DoWhy when observational data is all you get. Statistical maturity beats model-zoo knowledge here. Roles cluster at scale-ups and platforms with real traffic, where a one-percent effect pays several salaries.