EB-1A for data scientists and machine learning engineers

Machine learning is unusual among industry fields: its normal working artefacts are the ones the criteria were written to recognise. Papers at refereed conferences, citation counts, programme committee service, and society fellowships all exist natively in the field.

Authorship: conferences count

The criterion asks for "authorship of scholarly articles in the field, in professional or major trade publications or other major media", and the guidance defines what a scholarly article is before giving its examples:

"As defined in the academic arena, a scholarly article reports on original research, experimentation, or philosophical discourse. It is written by a researcher or expert in the field who is often affiliated with a college, university, or research institution. Scholarly articles are also generally peer reviewed by other experts in the field of specialization. In general, it should have footnotes, endnotes, or a bibliography, and may include graphs, charts, videos, or pictures as illustrations of the concepts expressed in the article." USCIS Policy Manual, Volume 6, Part F, Chapter 2

and:

"Examples of scholarly article authorship include, but are not limited to: Publications in professionally-relevant peer-reviewed journals; and Published conference presentations at nationally or internationally recognized conferences." USCIS Policy Manual, Volume 6, Part F, Chapter 2

That second example is the one that matters here. The field's most competitive venues are conferences, and their proceedings are published and refereed. A petition that treats conference papers as second-class because they are not journal articles is arguing against its own strongest evidence.

The guidance adds a note on affiliation: the definition says a scholarly article is "written by a researcher or expert in the field who is often affiliated with a college, university, or research institution". Often, not necessarily. Industry authorship is not excluded by the definition.

Citations, and where they are weighed

Citations do not appear in the ten criteria at all. They appear in the guidance twice: as evidence that a contribution is of major significance, and at the final merits stage.

"The petitioner provides evidence demonstrating that the total rate of citations to the person's body of published work is high relative to others in the field" USCIS Policy Manual, Volume 6, Part F, Chapter 2

The comparison is the whole point: high relative to others in the field, not high in absolute terms. A field that cites heavily produces high counts for ordinary work, which is why the record has to supply the comparison set rather than the number alone. Building a citation record covers how to construct that comparison, and defining your field covers why the denominator you choose decides the answer.

The guidance also names journal ranking as relevant, referring to evidence "regarding the journal's impact factor" and noting that "there may be particular prestige or acclaim associated with publication in such journals, especially if the person is the most significant contributor to the published article, a senior author, or the sole author of the article."

Author position matters. On papers with long author lists this is worth addressing directly rather than leaving to the officer.

Memberships: the example is in the guidance

The guidance offers an example that is unusually on the nose for this field:

"Another possible qualifying example may include membership as a fellow in a scientific society dedicated to artificial intelligence if the membership is based on recognition of a nominee's significant, sustained contributions to the field of artificial intelligence, and a panel of current fellows makes the selection of new fellows." USCIS Policy Manual, Volume 6, Part F, Chapter 2

Note the conditions attached: recognition of significant, sustained contributions, and selection by a panel of current fellows. It is the selection process that does the work, not the name of the body. The guidance elsewhere warns that membership based solely on education, years of experience, or payment of a fee is a factor pointing the other way. Memberships that qualify covers what to document.

Judging

Programme committee service and reviewing for conferences map directly onto the guidance's examples, which include "Peer review of abstracts or papers submitted for presentation at scholarly conferences in the respective field" and "Peer reviewer for government research funding programs."

The requirement is completion, documented: the petitioner must show the person "actually participated in the judging of the work of others", not merely that they were invited. Review systems in this field are ephemeral, and the confirmations are easy to lose. The evidence-capture habit and documenting a judging seat years after the fact cover both ends of that problem.

Contributions: models, systems, and adoption

For people whose strongest work is a deployed system rather than a paper, the criterion still fits, but the significance argument has to be built from adoption rather than from novelty.

The guidance names the relevant evidence, including "Documentation that the person's original work was cited at a level indicative of major significance in the field" and "Patents or licenses deriving from the person's work or evidence of commercial use of the person's work", and it warns that funding, patenting or publication alone "will not necessarily establish, on its own, that the work is of major significance to the field."

Where a criterion genuinely does not fit an industry role, the comparable evidence provision is available, and the guidance gives an industry example: presentation at a major trade show as comparable to the scholarly articles criterion, where publication is not readily applicable. Comparable evidence covers the argument and how it fails.

The real risk: recency

The criteria are a threshold. The case is decided at the second stage, on whether the record shows sustained national or international acclaim, and in a field that turns over quickly a strong record from four years ago reads differently from the same record from last year.

Sustained acclaim covers what sustained means, and the final merits determination covers the stage where it is assessed. This is the part of an ML profile most worth attention, because the categories are usually already satisfied.

EB-1A for software engineers with no PhD covers the case where none of the academic artefacts exist. Which criteria your own record supports is a legal judgement for your attorney.

Sources

All accessed 1 September 2026.

1. 8 CFR 204.5(h)(3) and (h)(4), eCFR. The ten criteria and the comparable evidence provision.
2. USCIS Policy Manual, Volume 6, Part F, Chapter 2. The scholarly article definition and examples, the artificial intelligence fellowship example, the citation and impact factor discussion, and the judging examples, all quoted verbatim above.
3. USCIS, Employment-Based Immigration: First Preference EB-1.

The mapping of machine learning artefacts to criteria in this article is practical guidance, not a regulatory standard.

This article is not a substitute for advice from your attorney about your own case.

Frequently asked questions

Green Card Guide AI is not a law firm and this article is not legal advice. It describes evidence-building strategy in general terms; your attorney determines what belongs in your petition and how it's argued.