EB-1A for data scientists and ML engineers

Also searched as EB1A for data scientists and ML engineers.

EB-1A for data scientists and ML engineers. A shipped model is usually job duties. What counts: methods others used, conference judging. Free check.

Who this is for

Applied scientists, ML engineers, and data scientists whose methods or models are used outside one employer.

Evidence that usually matters

Original contributions: a method, benchmark, or model others cite or deploy independently
Scholarly articles or peer-reviewed workshop papers with a real venue
Judging: conference or journal review in ML or statistics
Critical role on a product or lab that is distinguished in the field

What often does not hold up

Training a production model for one bank or product team is usually job duties
Kaggle rank without independent significance in the field
A citation count with no story of what the field did with the work

There is no separate EB-1A for machine learning

USCIS does not have an ML visa. The ten boxes are the same as for any sciences file. A production model at one employer is usually ordinary professional work. What tends to count is a method, benchmark, or system that independent groups used without you in the room, plus judging or papers that circulate in the field.

You can self-petition. No employer sponsor. No PhD requirement. Final merits still asks whether the whole record shows sustained acclaim at the top of the field, not only that a model shipped.

How the ten boxes usually look

Files that look plausible lean on original contributions, judging, scholarly articles, and sometimes a critical role. Kaggle, an internal model launch, and a Staff title are background. They are not, by themselves, three met criteria.

Original contributions: a method, loss, dataset, or system others adopted or cited independently
Judging: completed reviews for NeurIPS, ICML, ICLR, or a journal, with the invitation and proof you finished
Scholarly articles: a real venue in ML or statistics, not an internal wiki or a Medium post
Critical role: distinguished lab or product, and proof the result turned on your work
Awards and press: national or field-level, about you, not the company’s launch blog

What usually fails at final merits

One employer. One model. Letters from the same team. A citation count with no independent use. That can look busy on a resume and still look ordinary to an officer. If the free check comes back thin, read EB-1A vs NIW before you buy a long plan.

EB-1A is a self-petition. You can be in any country. The statute names five fields (sciences, arts, education, business, athletics). Meeting three criteria is only Step 1. Final merits still has to show sustained acclaim.

This is not a law firm and does not file petitions. USCIS decides each case on its own record.