EB-2 NIW for AI and Machine Learning Professionals

Every waiver runs on the same three-part test from Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016) — nationally important work, someone positioned to advance it, and a balance that favors waiving the job offer. There is no separate STEM test. What the guidance adds is a set of specific evidentiary considerations that make certain showings easier to reach inside that framework.

Officers identify a critical and emerging technology field by considering "governmental, academic, and other authoritative and instructive sources, and all other evidence submitted by the petitioner." USCIS Policy Manual, 6 USCIS-PM F.5(D)(4). Importance to competitiveness or security has content of its own: the guidance points to records showing that an endeavor will help the United States stay ahead of strategic competitors or current and potential adversaries. That vocabulary is available to AI and ML endeavors, and it has to be earned with sources.

The industry setting is not itself a problem. Many proposed endeavors that aim to advance STEM technologies and research, in academic and industry settings alike, have both substantial merit in relation to U.S. science and technology interests and implications broad enough to demonstrate national importance. USCIS Policy Manual, 6 USCIS-PM F.5(D)(4).

What is read differently is work whose benefit stops at the employer. Benefits to a specific employer alone are not sufficiently relevant to national importance, even where the employer has a national footprint. Applied ML that improves one company's ranking, pricing, or fraud models sits inside that rule — as does the chapter's closing illustration, a software engineer adapting an employer's code for clients, who will have difficulty showing national importance without broader impacts backed by specific evidence.

The routes out are named in the same paragraph: widespread interest in adopting or licensing the technology, a novel and important manufacturing or operational process, or an effect on how other companies develop similar technology. Methods, architectures, training and evaluation techniques, safety and interpretability work, benchmarks, and released infrastructure can all show the third — that others build differently as a result. Model-serving work done entirely inside one product, however senior, generally does not.

A degree does not establish national importance; the first prong is about the endeavor. It counts on the second and third. USCIS treats an advanced degree — particularly a Doctor of Philosophy — in a STEM field tied to the proposed endeavor and related to work furthering a critical and emerging technology, or another STEM area important to U.S. competitiveness or national security, as an especially positive factor weighed alongside the rest of the evidence. USCIS Policy Manual, 6 USCIS-PM F.5(D)(4). The reasoning is specific: doctoral holders have scientific knowledge in a narrow STEM area, and the officer then asks whether that specific area relates to the proposed endeavor. The link between what you researched and what you propose to do next is load-bearing. Theoretical work is expressly fine, since a theoretical area may still further U.S. competitiveness or national security as the proposed endeavor describes it.

The limit appears in the same section: a degree in and of itself is not a sufficient basis to find that a person is well positioned to advance the proposed endeavor. On the third prong the guidance treats a combination as a strong positive factor — an advanced STEM degree, especially a PhD; work furthering a critical and emerging technology; and being well positioned to advance that endeavor — with the benefit especially weighty where the endeavor has the potential to support U.S. national security or enhance U.S. economic competitiveness. Letters from interested government agencies are not required, but they can be relevant to all three prongs.

Which area, specifically, and how do you know it is critical? The record does the identifying, from governmental, academic, and other authoritative and instructive sources — not your own characterization of the field. USCIS Policy Manual, 6 USCIS-PM F.5(D)(4).

Does the endeavor reach past your employer? The officer asks whether the individual endeavor stands to have broader implications, such as for a field, a region, or the public at large.

Does your training connect to what you propose to do? The PhD is credited when it sits in a STEM field tied to the proposed endeavor. A gap there costs the strongest second-prong factor available. Every petition is decided case by case.

The technical area comes before the job description: which narrow area you work in, what authoritative sources say about it, and whether the sourcing actually supports a competitiveness or security framing. If it does not, we build on a different theory instead of stretching that one.

The endeavor gets separated from the employer's product early. For applied ML roles that separation usually decides the case, and it is far cheaper to answer before an exhibit set has been built around a job description.

We keep the PhD in its place, and we say so when the record is not there yet. Sometimes a paper or an adoption event six months out changes the case materially, and filing before it is a false economy.