All prompts
Draft
Build an H-1B specialty occupation argument
Turns a real job (actual duties, actual hours, actual org chart) into a duty-by-duty degree mapping, instead of the impressive-sounding job description that draws the RFE.
Your prompt
2
Pressure-test it
3
Go deeper
Before you run it
What to gather first
Watch for
What comes back
See an example of what you’ll get
Position. Vantia Financial Technologies seeks a Data Scientist II; the position requires a bachelor's or higher in Statistics, Applied Mathematics, or Computer Science with graduate-level machine-learning coursework.
| Duty | % | Body of knowledge | Degree field | Evidence |
|---|---|---|---|---|
| Builds and validates gradient-boosted churn models on 40M-row event data | 35 | Supervised learning, regularization, cross-validation, class-imbalance correction, calibration | Statistics / CS (ML) | Ex. C-2 model documentation; Ex. C-4 validation report |
| Designs A/B protocols; computes power and minimum detectable effect | 20 | Experimental design, hypothesis testing, sequential testing and alpha spending | Statistics | Ex. C-6 test plan, Q1 pricing experiment |
| Owns the feature store schema | 15 | Data modeling, leakage prevention, point-in-time correctness | CS / Statistics | Ex. C-7 schema and review notes |
| Presents findings to the pricing committee | 15 | *Carries no weight*; communicating results is not itself degree-requiring | none | none |
| Mentors two analysts | 15 | *Carries no weight* as pleaded; no technical content specified | none | none |
Aggregate. Duties requiring the named degree account for 70% of the role. The remaining 30% is presentation and mentoring, and the letter should not lean on it.
Argument (criterion: degree normally required). The first duty is not "analyzing data." It is choosing among gradient-boosting configurations, correcting for a severely imbalanced churn label, and calibrating output probabilities that feed a live pricing decision, the sequence taught in the graduate machine-learning and statistical-learning coursework Ex. C-2 reflects. [FIGURE NEEDED - if the manager will give me the actual positive-class rate, it sharpens this sentence considerably. I am not putting a number into a USCIS filing that did not come from the employer.] An employee without training in regularization and calibration produces a model that looks accurate and prices wrong. [UNVERIFIED - confirm the criterion's current wording against 8 C.F.R. § 214.2(h)(4)(iii)(A); no regulatory text was supplied.]
Contradiction check.
- LCA Wage Level II against a letter describing independent judgment on production models. Level II is defensible here, but do not use "leads," "directs," or "sets strategy" anywhere in the letter. That reads as Level III work at a Level II wage.
- SOC 15-2051 (Data Scientists) matches duties 1–3 cleanly. Good fit, no change needed.
- Supervisor holds a Ph.D. in Statistics: cite this in the employer-practice paragraph.
- Open flag: the 2024 written job description was not provided. If it says "or equivalent experience," it undercuts the entire letter and must be revised before filing, not after.
---
Assumptions I made.
- The time allocations came from the hiring manager and not from the job posting [verify - they total exactly 100, which usually means someone reconstructed them after the fact].
- The criterion argued in the section above is worded as I have described it [UNVERIFIED - you supplied no regulatory text. That sentence is bracketed and must be replaced with the language you pull, not adjusted around it].
- The two prior holders' M.S. degrees are documented in personnel records Vantia can produce, not remembered by the manager [verify].
- The LCA reads Level II at SOC 15-2051 with an Austin worksite as certified, and Austin is where she will actually sit [verify - a worksite mismatch is its own RFE, and it is the cheapest one to avoid].
- Vantia has no other live posting for this or a comparable title [verify - the officer can run that search, and you cannot un-run it].
Where this is weakest. Two places. The mentoring duty at 15%: as pleaded it does nothing but dilute the aggregate. Ask the manager whether 15% is really the number. If most of it is code review of model pipelines, say so and the degree-requiring share goes up honestly; if it is genuinely 15% of unspecified mentoring, leave it and live with 70%. Do not move a percentage to make the letter read better. Second: the argument paragraph rests on a criterion I have not read. An officer comparing that sentence to the current text is the fastest way this letter loses credibility on everything else in it.
What only you can decide. *How many criteria to argue.* Build on the single strongest one and the record stays tight and internally consistent, with nothing to fall back on if the officer rejects it. Argue in the alternative and you cover more surface area, at the cost of a letter some officers read as hedging, and every added criterion is another assertion the employer lives with at a site visit and at extension. *The expert opinion.* Commission it now and it carries independent weight and lands with the petition: three to four weeks, real cost, and a thin opinion from someone with no relevant record hurts more than none. Hold it for an RFE and you spend nothing unless you need it, but you write it under a deadline with the officer's objection already framed against you. *The 2024 written job description.* If it says "or equivalent experience," revising it before filing is the honest fix and creates a document dated after the offer; filing without touching it preserves the record and hands the officer the contradiction. That call belongs to you and the client.
What would make this materially better. (1) The current text of the specialty occupation criteria and the Policy Manual chapter, pulled today: every legal statement in this draft is bracketed until they arrive, and no amount of work on the duty table substitutes for them. (2) The certified LCA as certified, not as requested; the entire contradiction check rests on what I was told it says. (3) The 2024 job description and any live Vantia posting for this title: the single fact most likely to sink this petition, and the only one I cannot see from here. (4) Personnel records evidencing the two predecessors' degrees. (5) The manager's account of what the mentoring actually involves.
Why this prompt is built the way it is
## Framework
1. **Duties, not titles.** The argument is built duty by duty. A title and a market survey prove nothing about this job.
2. **Body of knowledge, then coursework.** For each duty, name what the person has to know and where that is taught. A duty with no course of study behind it carries no weight and gets separated out.
3. **Aggregate honestly.** State what share of the role the degree-requiring duties account for. Under half is a finding, not a drafting problem.
4. **Specificity in the field.** "Business administration or related field" is the classic failure. If the requirement really is general, the argument runs through the concentration or the specific coursework.
5. **Contradiction check.** LCA wage level against claimed complexity, SOC code against actual duties, the supervisor's credentials, the employer's own postings for the same title.
6. **The employer's record does the proving.** Predecessors' degrees, the written job description, the competency framework, and an explicit note where none of that exists.
7. **Position separate from person.** Evidence the beneficiary is impressive is not evidence the job requires the degree.
8. **No regulatory text from memory.** The standard comes from the authority the lawyer supplies, or it gets marked unverified.