Last updated: October 2026. Checked against the USCIS Policy Manual (Volume 6, Part F, Chapter 5, current as of October 2, 2026), the BLS Occupational Outlook Handbook, the DHS STEM Designated Degree Program List, the White House Critical and Emerging Technologies list (August 2026) and America’s AI Action Plan (July 2025).
Yes, data scientists and machine learning engineers can qualify for a National Interest Waiver. A strong EB2 NIW data scientist petition does three things well. It names the right occupation, because the BLS handbook has a data scientist profile but no separate “machine learning engineer” profile. It links your degree, and often your thesis, to the endeavor in the way the Policy Manual describes. And it proves impact even though your models are usually proprietary. “AI is a national priority” is not an endeavor. USCIS’s guidance gives “certain critical areas of artificial intelligence” as an example of critical technology, not all AI work.
This guide is for data scientists, ML engineers, applied scientists and AI researchers. If your work is mostly reporting, dashboards or business analytics, start with our EB-2 NIW guide for data analysts. If you mainly build production software, see the software engineer guide. Here we focus on modelling and ML work. All example endeavors are hypothetical and labelled as such.
Key takeaways
- USCIS decides EB-2 eligibility first, then applies the three Dhanasar prongs (USCIS, January 2025).
- Pick the BLS occupation that matches your work: data scientists, computer and information research scientists, software developers, or statisticians. BLS says each typically needs at least a bachelor’s degree.
- The Policy Manual says officers should consider whether the narrow subject of your dissertation or thesis “relates to the proposed endeavor.” Use it.
- Tie the endeavor to a named AI subfield or national need, such as “distributed and privacy-preserving machine learning” or “adversarial robustness and AI security.”
- Proprietary models can be documented through adoption by others, patents, contracts, released methods and detailed expert letters.
EB2 NIW data scientist eligibility: which occupation are you in?
Under the USCIS Policy Manual (Volume 6, Part F, Chapter 5), the occupation through which you will advance your endeavor “must meet the definition of a profession”: one for which a U.S. bachelor’s degree or foreign equivalent is “the minimum requirement for entry” (8 CFR 204.5(k)(2)). It is “the petitioner’s burden” to show this, and USCIS’s EB-2 filing tips say: “The occupation underlying the endeavor is determinative.” Then the officer applies Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016): national importance, well positioned, and a balance of factors favoring the waiver.
The BLS Occupational Outlook Handbook has no separate “machine learning engineer” profile. ML work can fall under several of its occupations, such as computer and information research scientists or data scientists. ML job titles cover very different work, so match the occupation to what you actually do:
| If your ML or data work mainly involves… | Closest BLS occupation | What the BLS handbook says |
|---|---|---|
| Building predictive models and applying ML to large datasets | Data scientists | Typically need at least a bachelor’s degree in mathematics, statistics, computer science or a related field; the profile says they “use machine learning to classify or categorize data or to make predictions” |
| Inventing new algorithms, architectures or training methods (research scientist roles) | Computer and information research scientists | Typically need at least a master’s degree in computer science or a related field; duties include “techniques from data science and machine learning” |
| Deploying and scaling ML systems in production software (many ML engineer roles) | Software developers | Typically need a bachelor’s degree |
| Statistical methodology, experimental design, probabilistic modelling | Mathematicians and statisticians | Typically need at least a master’s degree in mathematics or statistics |
Choose one occupation and use it consistently in your petition letter, experience letters and endeavor statement. If you rely on a bachelor’s degree plus five years of progressive experience, the filing tips say that experience “must all have occurred after you obtained your degree and must relate either to the field in which you received your degree or the proposed endeavor.” A physics or economics graduate who moved into ML should show how those years connect.
How USCIS treats your STEM degree, thesis and Ph.D.
The Policy Manual has a section on advanced degrees in STEM. Four points matter most for data scientists:
- How STEM is defined. The NIW rules do not define STEM, but officers “may refer to” the STEM OPT definition. The DHS STEM list (last updated July 22, 2024) includes Artificial Intelligence (11.0102), Computer Science (11.0701), Statistics, General (27.0501), Computational Science (30.3001), Data Science, General (30.7001) and Bioinformatics (26.1103). Check the DHS Study in the States page for the current version.
- Your thesis can carry weight. The Policy Manual notes that “doctoral dissertations and some master’s theses concentrate on a particularized subject matter,” and that officers “should then consider whether that specific STEM area relates to the proposed endeavor.” For example, a thesis on graph neural networks connects more directly to a fraud-network detection endeavor than a thesis on image generation would.
- A Ph.D. helps but is not required. An advanced STEM degree, “particularly” a Ph.D., tied to the endeavor and to critical technology work is “an especially positive factor” for prong 2. But “a degree in and of itself” is not enough to show you are well positioned.
- The prong 3 combination. USCIS treats three facts together as “a strong positive factor”: an advanced STEM degree (especially a Ph.D.), work “furthering a critical and emerging technology or other STEM area important to U.S. competitiveness,” and being well positioned to advance the endeavor.
Want to know whether your degree, thesis and current role line up? Request a free EB-2 NIW profile evaluation. Our petition team will review your background and suggest how the occupation and endeavor could be framed.
Linking a data science proposed endeavor to U.S. AI priorities
Footnote 93 of the Policy Manual says critical and emerging technologies “may include, but are not limited to, certain critical areas of artificial intelligence or quantum information science.” Note the word “certain.” To identify those areas, officers consider “governmental, academic, and other authoritative and instructive sources.” Two current White House documents are the most direct:
- The Critical and Emerging Technologies list. Appendix A of the National Security Science and Technology Strategy (August 2026) lists “Artificial intelligence (AI) and autonomy” subfields. These include “interpretability and control,” “adversarial robustness and AI security,” “distributed and privacy-preserving machine learning,” “perception, recognition, and sensor fusion” and “foundation models, including large language, multimodal, and world systems.” Other areas include “materials by design utilizing AI” and “high-performance supercomputing, including for AI applications.”
- America’s AI Action Plan (July 2025). The plan says “many of America’s most critical sectors, such as healthcare, are especially slow to adopt” AI. It calls high-quality data “a national strategic asset” and notes that “the inner workings of frontier AI systems are poorly understood.” It also has sections on AI-enabled science and AI evaluations.
A matching subfield helps with substantial merit and with the prong 3 factor, but it does not prove national importance on its own. The Policy Manual says “benefits to a specific employer alone, even an employer with a national footprint, are not sufficiently relevant.” A model built “for use or sale by a given company” falls short if it benefits “the company or its clients alone.” What helps is evidence of “widespread interest in adoption or licensing of the technology,” or of how it “stands to impact the development of similar technology by other companies.” National importance can also come from public health or economic effects outside the list, so do not force your work into a list it does not fit.
Hypothetical weak vs. strong data science endeavors
All examples below are hypothetical, written for illustration only. They are not client cases and do not predict any outcome.
| Weak framing | Stronger hypothetical framing | Official hook | Evidence it would need |
|---|---|---|---|
| “Build recommendation models to increase engagement on my employer’s app” | Develop federated learning methods that let several hospitals train early-warning models without sharing patient records | “Distributed and privacy-preserving machine learning”; AI Action Plan on slow healthcare adoption | Pilot results at more than one hospital, clinical partner letters, a released method or paper |
| “Fine-tune large language models for clients’ customer-service chatbots” | Create open evaluation methods that test whether language models used by public agencies give robust, reliable answers | “Interpretability and control”; “adversarial robustness and AI security” | Benchmark adoption by outside teams, citations or talks, agency or research-center interest |
| “Use machine learning to improve my company’s demand forecasts” | Build ML surrogate models that speed up screening of battery materials for U.S. manufacturers | “Materials by design utilizing AI”; AI-enabled science | Lab or manufacturer collaborations, published results, licensing or pilot agreements |
The stronger versions name a problem, users outside the employer and a route for the work to spread. Each also depends on evidence. If you cannot document a claim, narrow the endeavor. For structure and wording, see how to write a proposed endeavor.
Evidence for an EB2 NIW data scientist petition when your models are proprietary
Most production models belong to the employer and cannot be shared. The Policy Manual’s prong 2 list “is not meant to be a checklist,” and it warns that “unsubstantiated claims would not meet the petitioner’s burden of proof.” These items fit ML work:
| Evidence type (Policy Manual) | How a data scientist can document it |
|---|---|
| “Patents, trademarks, or copyrights developed by the person” | Patents or applications on model architectures or methods, with evidence of why they matter to the field |
| “Documents showing technology that the person invented, or contributed to inventing, and how others use that technology” | Open-source libraries or released models with documented outside users, integrations or forks |
| “Contracts with companies using products that the person developed” | Customer, licensing or pilot agreements for a model you built (redacted where needed) |
| Published articles, citations, or “positive discourse around, or adoption of” the work | Workshop or conference papers, preprints, technical blog posts by others discussing your method |
| Letters from government agencies or quasi-governmental entities | A public agency or federally funded research center that uses or evaluates your work |
| A plan describing how you will continue the work in the United States | Datasets, partners, milestones, and how results will be released |
Internal results still help when presented carefully. A supervisor’s signed letter can explain the problem, your role, the model’s purpose and the measured outcome compared with a baseline, without disclosing weights or data. Results carry more weight when another organization confirms or relies on them. The Policy Manual says letters are persuasive when experts with “first-hand knowledge” give “specific examples” and “are supported by other independent evidence.” Our guide to evidence without publications covers more options.
Prong 3 for AI and machine learning endeavors
The prong 3 factors include whether labor certification is impractical, the benefit of your work “even if other U.S. workers were also available,” urgency, and “unique knowledge or skills exceeding the minimum requirements standard.” Where the STEM strong positive factor applies, USCIS says the benefit is “especially weighty where the endeavor has the potential to support U.S. national security or enhance U.S. economic competitiveness,” or when letters from interested government agencies support it. One argument fails on its own: “evidence of a national labor shortage in the person’s occupation would not, by itself, satisfy this third prong.” Demand for AI talent is not a prong 3 argument; your specific endeavor is.
Common pitfalls for data scientists and ML engineers
- “AI” as the endeavor. Saying you will “advance artificial intelligence” names a field, not an endeavor.
- A job title with no occupation. “ML engineer” without a link to a BLS occupation and its entry requirements.
- Metrics without context. Accuracy or AUC numbers with no baseline and no explanation of who benefits outside the company.
- A thesis that does not connect. Claiming the Ph.D. factor when the dissertation topic has no link to the endeavor.
- Missing initial evidence. Since August 5, 2026, USCIS may deny a request without first issuing an RFE when required initial evidence is missing.
Frequently Asked Questions
Can a machine learning engineer file an NIW?
Yes, under the same rules. Because BLS has no ML engineer entry, describe the work and map it to data scientists, computer and information research scientists or software developers.
Do I need a Ph.D. for an AI NIW green card?
No. A master’s degree, or a bachelor’s plus five years of progressive experience, meets the advanced degree threshold. A Ph.D. tied to the endeavor is an especially positive factor, not a requirement.
Does any AI work count as a critical and emerging technology?
No. USCIS refers to “certain critical areas” of AI. Point to the specific subfield your endeavor advances and show evidence that the work reaches beyond your employer.
How is this different from a data analyst NIW?
The legal test is the same. Data scientists usually build models rather than reports, so the occupation, the thesis link and the AI priorities differ. Analysts should read the data analyst guide.
Is premium processing available?
Yes, for E21 NIW petitions. Form I-907 costs $2,965 (G-1055, edition 10/16/26), and USCIS must take action within 45 business days. That action can be a request for evidence rather than a decision (USCIS premium processing page).
Skycap prepares complete I-140 petition packages for data scientists and ML engineers filing on their own, with installment options on our plans page. To see how your modelling work could be framed, start with a free profile evaluation.
Related guides
This article is general information, not legal advice. Skycap Consultancy is a petition-preparation consultancy, not a law firm. Requirements and fees change; check the linked USCIS pages before filing.
