Machine Learning Engineer
Reported across 10 occupational codes in this source.
Additional reported codes account for the remaining 11 records.
We analyzed an official U.S. Department of Labor disclosure file to see how employers classified the titles they submitted. The result explains a core problem with salary websites: a job title often is not a precise government wage category.
These are not job openings or labor-market salary observations. The source is a selected immigration-program filing dataset that includes certified, withdrawn and denied cases. Its records do not represent the full U.S. workforce, and employers may classify the same title differently.
Consider Machine Learning Engineer. Among the FY2026 Q4 records bearing that exact normalized title, employers used ten different SOC codes. Most matching records were classified as Software Developers or Data Scientists, and a smaller number used still other categories. A salary site that silently maps the title to just one occupation would hide that ambiguity.
The pattern is not confined to AI. Senior Technical Program Manager and Software Engineer were also recorded under multiple codes. Not every variation is a meaningful professional distinction; some may be coding differences, old classifications or errors. The data shows that classification varies, not that all classifications are equally correct.
Reported across 10 occupational codes in this source.
Additional reported codes account for the remaining 11 records.
Reported across 24 occupational codes in this source.
Additional reported codes account for the remaining 124 records.
Reported across 4 occupational codes in this source.
Additional reported codes account for the remaining 1 records.
Reported across 8 occupational codes in this source.
Additional reported codes account for the remaining 10 records.
Reported across 2 occupational codes in this source.
Reported across 15 occupational codes in this source.
Additional reported codes account for the remaining 37 records.
A person looking up “Senior Technical Program Manager salary” is asking about a specific type of responsibility and seniority. A BLS estimate for Computer and Information Systems Managers, Project Management Specialists, or another occupation can provide context, but none automatically measures pay for that particular title. The occupational category may mix junior and senior workers, different industries and different job scopes.
AvgPay therefore treats real-world title identification as distinct from occupational wage statistics. Where we have source-backed role-level detail, specific role guides explain what the work involves. Where a title is ambiguous, title search displays plausible classifications rather than assigning one silently. Our role comparison tool examines responsibilities and skills directly, while keeping salary evidence tied to its actual source population.
We inspected the official FY2026 Q4 Labor Condition Application disclosure spreadsheet, excluded blank worksheet-formatting rows, standardized submitted job-title casing and punctuation for comparison, and grouped records by exact normalized title and recorded seven-character SOC code. We then counted entries under each title-code combination. We did not use any wage fields, model job demand, scrape private profiles, or infer an individual's compensation.
Case statuses in the source include Certified (106,746), Certified - Withdrawn (10,127), Withdrawn (1,924), Denied (410). This is an administrative dataset for certain immigration programs: multiple records can relate to a single employer or position, and it cannot be generalized into national title demand. Some records use legacy or questionable SOC codes; we retained the reported codes without declaring them correct.
Original agency: U.S. Department of Labor, Office of Foreign Labor Certification, Labor Condition Application public disclosure data
Release: LCA_Disclosure_Data_FY2026_Q4.xlsx, FY2026 Q4. Audited 2026-11-21.
Source SHA-256: 525a090def156e650a33797dde22bcb2307f51bf6b1b46a52ef84cb01af67c03
Streaming audit of supplied official Excel release, ignoring empty formatting rows, normalizing job-title whitespace/punctuation/case, and grouping by exact normalized employer-submitted title plus recorded SOC. Counts are disclosure rows, not unique workers or active vacancies. No wage fields were used. All mapped SOC codes from the filing are retained without assuming the employer selected the correct statistical occupation.
Read the official Department of Labor disclosure data or our research methodology. This analysis should not be used to draw conclusions about prevailing pay, worker shortages, hiring probabilities, wage premiums, or employment trends.