The strongest ML engineers on the market often cannot check the box your ATS requires. Their GitHub is in your dependency tree; their résumé never reached a human.
They are shipping at research labs and startups abroad, contributing to the open-source projects your stack already uses, and answering “yes” to the sponsorship question on every U.S. application — which is why you have not seen them.
The O-1 standard is extraordinary ability, evidenced by things a hiring manager can actually read:
Candidates on O1DMatch arrive with profiles, evidence summaries and O-1 readiness indicators — the record is organized before you ever see the person.
Timing: this year's H-1B cap closed on July 17, 2026 and the next registration window is approximately March 2027. The O-1 files year-round, and premium processing returns a decision in 15 calendar days. AI and deep-tech teams hire this profile constantly; the O-1 evidence record in ML is unusually legible because so much of the work is public.
The fastest way to see whether this pool holds your next machine learning engineer: paste the listing. A person replies with matches, or with an honest "no match yet."
