Career Strategy
O-1A High Salary Criterion: Data Science and Machine Learning Compensation Benchmarks in 2026
Data scientists and machine learning engineers earn some of the highest salaries in tech, but constructing a persuasive high-salary exhibit still requires the right comparator data and precise framing. Here is how to build a salary exhibit that holds up to USCIS scrutiny in 2026.
The high-salary criterion and why data scientists are well-positioned
The O-1A high-salary criterion at 8 C.F.R. § 214.2(o)(3)(iii)(B)(8) requires evidence that the petitioner commands a high salary or other substantial remuneration relative to others in the field. For data scientists and machine learning engineers, this criterion is structurally favorable: the profession sits at the top of the U.S. compensation distribution across technical fields, with senior and staff-level practitioners at major technology companies earning total annual compensation that far exceeds the median for most STEM disciplines. The challenge is not reaching the high-salary threshold — most petitioners at the staff or principal level already do — but documenting the comparator data correctly so the exhibit withstands adjudicator scrutiny.
USCIS evaluates high salary against 'others in the field.' For data scientists and machine learning engineers, the field is the petitioner's specific subfield and career level — a staff machine learning engineer at a technology company is compared to other staff machine learning engineers, not to all data scientists or all software engineers. Defining the comparator group too broadly (comparing a senior ML engineer at a top-tier company against BLS data for all 'computer and information research scientists') artificially inflates the apparent differential. Defining it correctly and using comparator data that matches the petitioner's role, geography, and experience level produces a more persuasive exhibit that holds up to RFE scrutiny.
This article focuses on the evidence and framing strategies for the high-salary criterion specifically, drawing on publicly available compensation data sources and the regulatory standard. It does not address the other O-1A criteria for data scientists — those are addressed in separate articles on scholarly articles, judging, and original contributions. The audience is petitioners and practitioners preparing the compensation exhibit for a data science or machine learning O-1A petition.
Primary data sources for the high-salary comparison
The Bureau of Labor Statistics Occupational Employment and Wage Statistics (OEWS) survey is the regulatory baseline comparator. The relevant BLS SOC codes for data scientists and machine learning engineers are 15-2051 (Data Scientists), 15-1252 (Software Developers, Quality Assurance Analysts, and Testers — used when the role is primarily software engineering with ML components), and 15-1221 (Computer and Information Research Scientists — used for research-oriented ML positions). The BLS publishes 10th, 25th, 50th, 75th, and 90th percentile wages by SOC code, by industry, and by metropolitan statistical area. A petitioner whose total compensation — base salary plus cash bonus, not including unvested equity — exceeds the 90th percentile for their SOC code and metropolitan area has strong foundational documentation.
For technology industry roles, BLS data systematically understates compensation because it excludes equity, which constitutes a major fraction of total compensation at major technology companies. Supplementary data sources that capture equity compensation are therefore essential. The Radford Technology Survey (now part of AON's compensation database), the Levels.fyi salary database, the compensation data published by Glassdoor and LinkedIn Salary for specific companies and role levels, and internal compensation benchmarking reports from technology companies all provide evidence of equity-inclusive total compensation. Levels.fyi data is particularly useful because it documents base salary, cash bonus, stock grant value, and total annual compensation at the role level (E5, L5, etc.) and company level for major technology employers.
Specialized compensation reports from technology industry analysts — such as the Mercer Technology Compensation Survey, the Willis Towers Watson Technology Compensation Survey, and compensation reports published by the Computer Economics division of Avasant — provide aggregated industry benchmarks that carry more institutional authority than crowd-sourced platforms. Where these reports are available, they should be included as the primary comparators, with BLS data as a floor benchmark and Levels.fyi or similar as corroboration. The petition should explain the relationship between each data source, acknowledge what each captures and excludes, and show how the petitioner's actual compensation compares to each benchmark.
Defining the comparator group correctly
USCIS has issued guidance emphasizing that the high-salary comparator must reflect 'others in the field' — meaning others doing comparable work at a comparable career level, not the general workforce. For data scientists and machine learning engineers, the comparator group should be defined by role, seniority level, geographic market, and industry. A staff ML engineer at a San Francisco Bay Area technology company is properly compared to staff-level ML engineers in the San Francisco Bay Area technology industry, not to data scientists nationwide or to all software engineers in any location. BLS Metropolitan Area data for the San Francisco-Oakland-Hayward MSA, supplemented by Levels.fyi data filtered by company tier and role level, provides a well-defined comparator group.
The distinction between 'data scientist' and 'machine learning engineer' matters for comparator definition because the two titles command different compensation distributions in 2026. Machine learning engineers at major technology companies (identified as companies with more than 10,000 employees and a substantial technology revenue base) typically earn higher base salaries and larger equity grants than data scientists in analytical or business intelligence roles at comparable companies, because the ML engineering role requires production software engineering skills in addition to modeling expertise. A petition for an ML engineer should use ML engineer comparators, not generalized data scientist benchmarks, to avoid understating the differential.
The geographic dimension of the comparator group also requires precision. BLS Metropolitan Area data distinguishes among the San Francisco-Oakland-Hayward MSA, the San Jose-Sunnyvale-Santa Clara MSA, the Seattle-Tacoma-Bellevue MSA, and the New York-Newark-Jersey City MSA — each of which has a distinct compensation distribution for technology roles. A petitioner working in Seattle should not be benchmarked against San Francisco data, even though both are high-compensation markets. Using the correct metropolitan area data, or a multi-city average weighted by the petitioner's work location, ensures that the comparison reflects the actual labor market the petitioner competes in.
Documenting the petitioner's actual compensation
The petitioner's actual compensation must be documented concretely, not summarized by the attorney or stated in a declaration that USCIS has no means to verify. Offer letters and employment contracts showing base salary are the starting point. Annual bonus documentation — either a letter from the employer stating the target bonus percentage and the most recent actual payout, or compensation statements showing bonus payments received — addresses the cash variable compensation component. Equity documentation — restricted stock unit grant agreements, vesting schedules, and either the stock's current market price or the employer's trailing twelve-month average stock price — establishes the equity component's annualized value.
For petitioners employed at public companies, the stock price is verifiable from public market data, making the equity valuation straightforward. For petitioners at private technology companies, the equity valuation requires additional documentation: the most recent 409A valuation report (or the per-share price from the most recent funding round), the number of unvested RSUs or options, and the vesting schedule. The petition should explain the valuation methodology and its limitations, because USCIS adjudicators may scrutinize private company equity valuations. Where the private company has a well-documented valuation from a major venture capital round within the past twelve months, that valuation provides a credible basis; for earlier-stage companies with less certain valuations, relying more heavily on cash compensation and treating equity as supplementary context is often more conservative.
Total compensation statements from the employer — either on official letterhead signed by a compensation or HR officer, or from the employer's equity management platform — are among the most efficient documentation formats because they present base salary, bonus, and equity grant value in a single exhibit. Many major technology companies use equity management platforms that generate official grant statements showing the grant date value, vesting schedule, and current unvested balance. These documents, combined with a payroll or W-2 statement confirming base and bonus payments, provide a complete and verifiable compensation record.
Common problems in data science high-salary exhibits
The most common error in data science high-salary exhibits is using median compensation data as the comparator without establishing that the petitioner's compensation exceeds it by a meaningful margin. USCIS has not defined a specific threshold — 'high salary' is a relative term — but adjudicators routinely look for compensation at or above the 90th percentile for the comparator group. A petition showing that the petitioner earns 10 percent above the median does not make a persuasive case. The exhibit should show where the petitioner falls in the distribution — ideally at or above the 90th percentile — and include the source data that establishes the distribution.
A related error is comparing the petitioner's base salary against total compensation figures from survey data, or comparing the petitioner's total compensation against base salary figures from BLS. These apples-to-oranges comparisons produce misleading differentials that USCIS or, on appeal, the AAO may identify and discount. The exhibit should ensure that the petitioner's compensation figure (base, total cash, or total compensation) is compared against the same metric in the survey data. If the survey reports base salary and the petitioner's primary differentiator is equity, the exhibit should supplement the base-salary comparison with a separate total-compensation comparison using a source that captures equity.
Finally, high-salary evidence is most persuasive when paired with the other O-1A criteria rather than presented in isolation. A petitioner whose only strong criterion is high salary faces a more difficult adjudication because USCIS has indicated that salary alone does not establish extraordinary ability — it must be accompanied by evidence of achievement (awards, scholarly articles, expert recognition, or significant original contributions) that explains why the salary is high. The compensation exhibit is strongest when it reinforces a narrative of extraordinary achievement rather than serving as a standalone claim.
Constructing and auditing the high-salary exhibit
A well-constructed high-salary exhibit for a data science O-1A petition includes: the petitioner's offer letter or employment contract showing base salary; annual bonus documentation (target percentage and most recent actual payment); equity grant agreements and a supporting stock price or 409A valuation; a total compensation calculation prepared by the attorney drawing on these documents; BLS OEWS data for the relevant SOC code and metropolitan area, with the petitioner's base salary plotted against the percentile distribution; and supplementary compensation survey data from Radford, Mercer, or Levels.fyi (filtered to role level and geographic market) showing total compensation comparators. Each exhibit should be introduced with a brief explanatory cover, and the attorney's cover letter should walk through the comparison explicitly.
Before filing, audit the exhibit by stress-testing the comparison from a hostile reader's perspective. Would an adjudicator who picks up a narrower comparator group than intended still conclude the salary is high? If the answer is no, the exhibit needs strengthening. Check that the petitioner's compensation figure and the comparator data are on the same basis (base-to-base or total-to-total). Verify that the geographic market for BLS data matches the petitioner's actual work location. Confirm that the Levels.fyi or survey data is filtered to the appropriate role level and company tier. Address any obvious gaps — such as an equity component that represents the majority of the petitioner's compensation but lacks valuation documentation — before they become RFE fodder.
The high-salary criterion for data scientists and machine learning engineers in 2026 is achievable for practitioners at the staff or principal level at major technology companies. The evidentiary challenge is documentation and framing, not reaching the threshold. A compensation exhibit that uses the correct comparator group, presents the petitioner's compensation on the same basis as the survey data, and maps the result to the appropriate percentile of the distribution gives the adjudicator everything they need to find the criterion satisfied. That specificity — not a general assertion that data science pays well — is what turns a procedurally adequate exhibit into a persuasive one.
What we typically gather for this kind of case
| Document | Where to source | Why it matters |
|---|---|---|
| Peer-reviewed publications | Web of Science / Scopus exports | Anchors original-contributions and authorship criteria |
| Citation analysis | Google Scholar profile + ESI top-1% data | Quantifies major significance in the field |
| Salary benchmark | BLS OEWS for SOC code + locality | Documents high-salary criterion at 90th-percentile or above |
| Critical-role letters | Direct supervisor + program director | Establishes role's importance, not just title |
What we see go wrong, again and again
- 01Treating extraordinary ability as a credentials checklist rather than a story of field-wide impact.
- 02Submitting bibliometric data (h-index, citation counts) without explaining what makes those numbers high relative to peers in the same sub-field.
- 03Relying on letters from collaborators or co-authors rather than independent experts who can speak to influence.
See if you qualify
Lando reviews your background against the O-1 visa criteria and tells you honestly where you stand. Free, no commitment.