EB-2 National Interest Waiver Petition

This petition seeks classification of Dr. M. as a member of the professions holding an advanced degree (Ph.D. Computer Science, Carnegie Mellon, 2020) and requests, under Section 203(b)(2)(B) of the Immigration and Nationality Act, a national-interest waiver of the labor certification requirement.

Dr. M.'s proposed endeavor — developing interpretable foundation-model safety methods for high-stakes US deployment domains — sits at the intersection of three formally identified U.S. national priorities: the National AI Initiative Act of 2020 (Pub. L. 116-283 §5301), Executive Order 14110 on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (Oct. 30, 2023), and the NIST AI Risk Management Framework v1.0 (Jan. 26, 2023).

As demonstrated below, Dr. M.'s proposed endeavor (a) has substantial merit and national importance, (b) is one Dr. M. is well positioned to advance, given a publication record of 32 peer-reviewed papers (2,418 citations, h-index 18), two issued U.S. patents, and a NeurIPS 2023 Spotlight Award, and (c) on balance, would benefit the United States to waive the requirements of a job offer and labor certification.

The United States faces a documented and accelerating gap between AI capability and AI safety capacity. The 2024 Department of Labor Bureau of Labor Statistics occupational outlook identifies AI/ML research as one of the top-five high-skilled occupations with persistent unfilled demand, and the National Science Foundation's 2024 AI Talent Report concludes that the U.S. produces fewer than 1 in 4 of the foundation-model safety researchers needed to meet the goals of EO 14110.

The Office of Science and Technology Policy's 2024 R&D priorities memorandum (M-24-04, July 1, 2024) explicitly elevates "trustworthy AI" — including interpretability, robustness, and adversarial resilience — to a top-tier funding category for FY2025. The CHIPS and Science Act (Pub. L. 117-167, §10391) authorizes the National AI Research Resource specifically to expand "research on responsible and trustworthy AI." Dr. M.'s proposed endeavor falls squarely within these formally articulated priorities.

[…full section continues with citations to NSF SaTC program announcements, DARPA AI Forward agenda, and the NIST GenAI Profile…]

Under Matter of Dhanasar, 26 I&N Dec. 884 (AAO 2016), the first prong asks whether the proposed endeavor has "both substantial merit and national importance." Substantial merit may be demonstrated in a range of areas, "including but not limited to business, entrepreneurialism, science, technology, culture, health, or education." Id. at 889.

Dr. M.'s endeavor satisfies substantial merit on the science-and-technology branch unambiguously. AI safety has been formally identified as a U.S. priority across three branches of government: the legislative (CHIPS and Science Act §10391), the executive (EO 14110, OSTP M-24-04), and the agency level (NIST AI RMF v1.0, NSF SaTC, DARPA AI Forward).

National importance, under Dhanasar, is satisfied where the endeavor "has broader implications" beyond the petitioner's immediate employer. AI/ML safety methods are inherently public goods: published interpretability techniques propagate across the entire foundation-model ecosystem, including federally regulated deployment domains (healthcare under HIPAA, financial services under the Fair Credit Reporting Act, and critical infrastructure under CISA's 2024 secure-by-design guidelines).

[…full section continues with paragraph-level evidence pinning each claim to a specific federal document…]

Prong 2 asks whether Dr. M. is "well positioned to advance the proposed endeavor." Dhanasar, 26 I&N Dec. at 890. The AAO has identified several non-exhaustive factors: education, skills, knowledge, record of success, and a model of execution.

Dr. M.'s record of independent recognition includes:

• 32 peer-reviewed publications (11 first-author), 2,418 cumulative citations, h-index 18, i10-index 24 — figures that, in our analysis of recent AAO decisions, exceed the median for approved EB-2 NIW petitions in the AI/ML field.

• A NeurIPS 2023 Spotlight Award, granted to fewer than 3% of submissions to the world's most cited AI conference. NeurIPS Spotlight selection has been recognized by AAO panels as evidence of peer recognition at the international level (see, e.g., Matter of [redacted], JAN 2024 AAO 0142).

• Two issued U.S. patents (US 11,XXX,XXX and US 11,YYY,YYY) covering foundation-model interpretability techniques, both currently licensed by U.S. industry partners.

• Independent media recognition: featured in MIT Technology Review (March 2024) and TechCrunch (October 2023), both naming Dr. M.'s research group's work as among the year's most consequential in AI safety.

[…full section continues with a per-criterion evidence map, citing Dr. M.'s Google Scholar metrics directly and 4 of the 5 strongest similar-case approvals from our AAO dataset…]

Prong 3 asks whether, "on balance, it would be beneficial to the United States to waive the requirements of a job offer and thus of a labor certification." Dhanasar, 26 I&N Dec. at 890.

Several factors cut decisively in favor of waiver here:

(1) Impracticality of labor certification. The Department of Labor's PERM data for FY2023–2024 (cited in the 2024 NSF AI Talent Report) shows that AI/ML research positions at Dr. M.'s seniority level have median PERM-equivalent processing exceeding 18 months — a delay incompatible with the urgency articulated in EO 14110 §4.4 (90-day timelines for NIST AI RMF rollouts).

(2) Public-good nature of the work. Dr. M.'s research outputs — interpretability methods, open-source toolkits, and peer-reviewed safety findings — propagate freely across the U.S. AI ecosystem, regardless of any specific employer's immediate use. Conditioning Dr. M.'s contribution on a single employer's labor certification would substantially constrain the diffusion of these public-good outputs.

(3) Documented national shortage. The 2024 NSF AI Talent Report concludes the U.S. produces materially fewer foundation-model safety researchers than national priorities require.

[…full section continues, closing with explicit waiver-balancing language drawn from approved AAO precedents…]

For the foregoing reasons, Dr. M. respectfully requests classification as a member of the professions holding an advanced degree under INA §203(b)(2)(B) and grant of a national-interest waiver of the labor certification requirement.

Dr. M.'s proposed endeavor — interpretable foundation-model safety for high-stakes U.S. deployment domains — has been shown to (a) hold substantial merit and national importance, anchored in three branches of explicit U.S. policy; (b) be one Dr. M. is well positioned to advance, supported by a peer-recognized record of publications, patents, and awards exceeding the median for approved EB-2 NIW cases in the field; and (c) on balance, benefit the United States to waive labor certification, given the documented urgency, the public-good character of the work, and the formally identified national shortage.

The accompanying exhibits include [list], in support of each Dhanasar prong.