Or peripheral to it, e.g. data engineering for pre-training data or a joint DE/ML role, etc. Sometimes the scale (petabyte) or specific stack experience is a barrier.