Own a production CTR/CVR prediction model end-to-end — modeling, eval, feature pipelines, online experimentation, and post-launch ops. By month six, you'll own a meaningful slice of NEXT's ML stack.
Hunt for the missing signals that move the needle: new data sources to log and ingest, derived and contextual features the current model doesn't yet see. On NEXT, most wins come from finding signals others missed — not from architectural cleverness.
Run the loop fast: design offline evaluation, ship to online A/B, read out in days, iterate. Diagnose the offline-online divergences when they show up — and they will.
Build the agentic tooling that automates parts of our experiment-debugging and signal-discovery workflow, both as a contributor and as a user.
Set technical direction. Decide what NEXT should bet on next quarter, not just execute on assignments. Bridge to the data and pipeline teams whose signals feed our models — most signal-hunting wins depend on getting those teams aligned.
Embrace the unglamorous parts: data-quality instrumentation, train/serve consistency in feature pipelines, slicing eval to find failure modes, and the careful experiment debugging that separates real wins from noise.
How Do I Know if the Role is Right For Me?
5+ years of machine learning experience with a track record of shipping production-grade models in business-critical environments. We don't filter on degrees.
Experience with data analysis
Experience working on large-scale prediction or decisioning systems — CTR/CVR, ranking, recommendation, personalization, or related.
