Pivot from t-shaped generalist DS into a frontier-adjacent specialty (eval / RLHF tooling, MLE infra, agent systems) and ship one named OSS artifact in it this year — this is the single sharpest defence against the AI automation overhang on classical data science.

Move from well-known employer to a frontier-AI lab or a hot AI infra startup — at this experience level, the employer-quality jump alone shifts comp and transfer cost by 3-5x.

Convert active visibility into 1-2 conference talks or a technical blog with real readership — passive 'active' visibility doesn't generate inbound from frontier recruiters; named talks do.