Outricher · Signal Expansion · Appendix to the visa research
From one filter to a targeting engine.
Expanding the A2 request across O-1 / EB-1A / NIW.
The client asked for a specific slice — Indian senior engineers, big-tech pedigree, EB-1A self-petition. That is one good cell of a much larger matrix. This appendix takes the request apart as a business + data analyst would: it adds the intent axis the request is missing, re-shapes the score the way USCIS actually adjudicates, adds a negative-search layer, and shows how the same machine extends to O-1 and NIW. Every signal is tagged by what our data can and cannot see.
2
axes the buyer asked for — want & can; the request encodes only one
3
visa routes, 3 different petitioner logics — one score mis-ranks two
6
free DB signals to add this round — biggest is the intent axis
0
fabricated fields — triage layer, never an eligibility verdict
Appendix to Visa Candidate Intelligence — O-1 / EB-1A / NIW. Analyst layer above the live A2 extraction pipeline; does not change that pipeline's contract. Grounded in current (2024–2026) USCIS regulation + a live signal inventory, 2026-06-16.
← Main research: Visa Candidate Intelligence — O-1 / EB-1A / NIW
01What was asked vs. what is wanted
The buyer's own brief names two axes: people who (1) urgently want an O-1 / EB-1 / NIW and (2) can get one on background. The A2 filter encodes only the second — and only one archetype of it. Three structural gaps follow.
Gap 1 — highest value
The intent axis is missing
The "want it urgently" half is not modeled. Visa keywords sit inside the qualification score as a small bonus, conflated with merit. Qualification alone is a cold list; an agency that sells now needs people actively trying to file. Almost free to compute from data we already pull.
Gap 2
One archetype, three visas
A2 is EB-1A only. O-1 needs a US petitioner/employer; EB-1A & NIW are self-petition. So "has a sponsoring employer" carries the opposite sign per visa. A single global score mis-ranks two of the three routes.
Gap 3
An additive box-counter
The scorer sums all signals into one number. USCIS uses Kazarian's two steps: count ≥3 criteria, then a "sustained acclaim / very top of field" merits judgment. A pure sum over-predicts — a Big-Tech senior lights up pedigree+salary+seniority with zero actual acclaim.
A2 still ships unchanged. It is a well-chosen operational slice for this round. Everything below adds layers around it — turning "a list of senior Indian engineers" into "a ranked, evidence-tagged, intent-sorted shortlist," and showing the path to the other two visas.
02The legal → signal matrix
Each USCIS criterion mapped to the strongest professional-profile proxy we can actually compute, tagged by where the data lives and how strong the proxy is. The "In A2 now?" column shows what the current pipeline already does and where the cheap wins are.
DB — native column / indexed table
DB-text — regex on text we pull
DB-derive — computed in Node
API — paid call
— not observable
O-1A & EB-1A — meet ≥3 of 8 / 10
| # | Criterion | Best professional-profile proxy | Source | Strength | In A2 now? |
|---|
| 1 | Awards & prizes | profile awards registry (121M) + name registry-match | DB | Medium | ✓ EXISTS +6 |
| 2 | Elite membership | "Fellow of IEEE/ACM", "National Academy" in headline/summary | DB-text | Medium | add |
| 3 | Press about the person | "featured in / as seen in" + links only | DB-text | Weak | flag-only |
| 4 | Judging others' work | "reviewer for", "program committee", "editorial board", "grant panel" | DB-text | Strong* | add — high value |
| 5 | Original contributions | patents (6.2M), adopted OSS (GitHub stars), "inventor of", standards | DB +text | Medium | partial (EXISTS) |
| 6 | Scholarly authorship | publications registry (74M) + "author" + Scholar link | DB | Strong | ✓ EXISTS +6 |
| 7 | Critical role at a distinguished org | senior title × name-brand company (pedigree join) | DB | Strong | ✓ pedigree+seniority |
| 8 | High remuneration | inferred salary band (~6–16% cov) → modeled band | DB | Weak | ✓ band +6/+4/+2 |
* "Judging others' work" is strong when stated — one of the most cleanly observable criteria, and currently missing from the score.
EB-2 NIW — Matter of Dhanasar, three prongs
| Prong | Test | Best professional-profile proxy | Source | Strength |
|---|
| Gate | Advanced degree (MS/PhD) or 5+ yrs progressive | degree (MS/PhD) + tenure | DB derive | Strong |
| 1 | Substantial merit & national importance | field ∈ Critical & Emerging Tech (§04) from industry+headline+skills | DB-text | Medium |
| 2 | Well-positioned to advance (incl. funding/adoption) | "raised $X", founder; employer funding via the funding-data link | DB-text DB | Med–Strong |
| 3 | Beneficial to waive labor cert | self-petition disposition (founder / independent / between-jobs) | DB-derive | Weak |
NIW's strongest cluster (Jan-2025 policy): advanced STEM degree (esp. PhD) + a CET field + funding/traction. Mostly DB-derivable — the highest-confidence path if we extend beyond A2's EB-1A archetype.
03The intent axis — the cheapest high-impact add
This is the buyer's "want it urgently" half — and the cheapest big win. Intent is a multiplier on qualification, never an additive term. A perfectly qualified profile with zero intent is a poor lead for an agency selling now; a moderate profile posting "preparing my EB-1A" is hot.
🎯
Profile names the visa explicitly
"O-1 / EB-1A / NIW / green card / self-petition / extraordinary ability" · DB-text · strength: strong
Rare but golden — the highest-intent signal possible. Pull it out of the qualification score into its own axis.
⚡
Open-to-work + recent role end
open-to-work flag + recent last-position end date · DB · strength: strong
Foreign national just laid off = status cliff = urgency. Not in A2 today; free to add.
⏳
OPT / STEM-OPT / H-1B pressure
visa-status mention + grad year (cap-gap / lottery) · DB-text + education dates · strength: strong
Often self-stated; otherwise modeled from grad date + visa-dependent origin.
📣
Recent posts about the visa process / RFEs
activities[] · API (free inside the 2000 full fetches already budgeted) · strength: strong
Recency-weighted — a 2026 post far outweighs a 2022 one. Capture when present; don't gate on it.
👥
Immigration groups / follows attorneys
groups[] · API-only, 1 credit · strength: medium
Bonus-if-present from the full profile fetch, never a separate spend.
For A2 this round: the DB-text intent signals + the open-to-work flag are free — emit them as a separate intent_score. The activities[] / groups[] signals come for free inside the 2000 full-profile API fetches already paid for — capture when present.
04National-importance field hooks (NIW Prong-1)
Grounded in the Jan-2025 USCIS NIW guidance referencing the federal Critical & Emerging Technologies (CET) list + the 2025 AI Action Plan. Mapped from industry + headline + skills. The field is necessary, not sufficient — USCIS credits the specific endeavor's impact, which professional data cannot show. Score the tier; don't let it alone promote a profile.
Highest
AI / ML · semiconductors & microelectronics · quantum information
High
biotechnology · cybersecurity / critical infra · clean & advanced energy · advanced manufacturing & robotics
Med–High
critical supply chains (pharma, rare materials, defense) · aerospace / autonomous / space · public health / vaccine R&D
Low
generic IT / web / SaaS / fintech with no CET tie — won't carry Prong-1 alone
05Reframed score — router → (gate × eminence) × intent
Replace the single additive number with a small structured model. Every piece is computable from signals we already pull — the change is in arrangement, not new infrastructure. The current weights become the eminence inputs almost unchanged.
STEP 0 CATEGORY ROUTER
has US employer / US-incorporated founder? → O-1A viable (needs a petitioner)
founder / independent / between-jobs / solo acclaim? → EB-1A / NIW (self-petition)
advanced STEM degree + CET field? → NIW preferred path
→ route to the visa that actually fits; never one global scale.
STEP 1 QUALIFICATION = criteria_gate (count ≥3 observable) [Kazarian step 1]
× eminence_score (continuous) [Kazarian step 2 proxy]
eminence = f(seniority, pedigree tier, breadth of independent signals,
sustained-over-years, patent/pub COUNT not just presence, followers pctile)
STEP 2 INTENT = behavioral urgency multiplier (§03)
FINAL RANK = Qualification × Intent — surface BOTH axes to the buyer, never just the product
with hard-excludes removed + soft-penalty flags annotated (§06)
O-1A
Petitioner REQUIRED
A US employer / agent is itself a gating signal. Favor candidates with a US role, a sponsoring employer, or a US-incorporated company they founded. Employer prestige doubles as Criterion 7.
EB-1A
Self-petition
De-weight "has a sponsoring employer" — the candidate is the petitioner. A profile that looks "stuck" employer-wise but has independent acclaim is a strong EB-1A lead.
EB-2 NIW
Self-petition
Employer irrelevant as sponsor. Weight the individual's funding / traction / adoption as Prong-2 evidence. PhD + CET field + funding is the highest-confidence cluster.
Why this beats the sum. The router stops mis-ranking self-petition vs. employer-sponsored profiles (Gap 2). gate × eminence separates "very top of field" from "senior but ordinary" (Gap 3). × intent re-ranks the qualified pool by who is trying to file now (Gap 1). Emit qualification_score, eminence_score, intent_score, criteria_met_count, visa_route per lead so the client can re-segment without a re-run.
06Negative search — looks qualified, is a bad lead
Two classes: hard excludes (remove) and soft-penalty + human-review flags (keep, annotate, rank down). The most valuable exclude — citizen / green-card holder — is also the least reliable from professional data, so it must be a flag, never a hard drop.
| Negative signal | Action | How we detect it | Reliability |
|---|
| Junior / intern / student / PhD-candidate, no record | Hard exclude | title+headline regex (A2 already does) + empty patents/awards/pubs and no leadership | High — cleanest exclude |
| 10+ yrs but zero eminence signals | Rank down | no acclaim, no leadership, single-employer → likely employer-sponsored EB-2/EB-3, not self-petition lane | Medium |
| No national-importance field (retail / hospitality / generic sales) | Rank down (NIW) | industry classification | High |
| Retiree / emeritus / "formerly" | Soft penalty | "retired"/"formerly" + last role ended years ago + stale activity | Medium |
| Likely US citizen / green-card holder | Flag, never auto-exclude | "US Citizen" in text, clearance roles, long US-only history w/ no foreign education | Low — false-confidence risk |
| Already on a stable long-term path (approved I-140) | Soft penalty if inferable | rarely stated | Low |
Two negatives NOT to apply blindly: employment gaps and job-hopping. For EB-1A/NIW these are often founder periods or independent-research stretches — penalizing them would drop exactly the self-petition profiles we want. For A2 specifically, the India-undergrad + US filter already makes the citizen/GC false-positive small (India-undergrad + US ≈ almost certainly a foreign national); the flag matters most the moment we extend beyond A2.
07What ships this round vs. later
A2, DB-only + the 2000 paid full-profile API fetches the client authorized. The cheap wins are the intent split and the two missing text criteria.
| Addition | Cost this round | Call |
|---|
Split intent_score out of qualification (DB-text + open-to-work) | Free | DO NOW — biggest value |
| "Judging / reviewer / editorial board" text signal (#4) | Free | DO NOW — strong, cheap, missing |
| "Fellow of IEEE/ACM" membership text signal (#2) | Free | DO NOW |
| Patent / publication COUNT (not just EXISTS) for eminence | Free | DO NOW |
Emit criteria_met_count + eminence_score + visa_route per lead | Free (re-arrange weights) | DO NOW — lets client re-segment |
| CET-field tier for NIW route | Free | If extending to NIW |
groups[] / activities[] intent signals | Free inside the 2000 fetches | Capture-if-present |
| External verification (Scholar citations, USPTO grants, press) | Out of DB + budget | DEFER — "attorney-to-verify" |
| Citizen / GC flag | Free but low reliability | Advisory flag only |
08Honesty boundary & a live legal wildcard
What professional data cannot show — to be stated plainly in the delivery so the buyer never resells this as an eligibility verdict (UPL + deceptive-practice + brand risk). These are exactly the artifacts USCIS adjudicates.
- Citation counts / publication impact — only that a publication exists
- Real salary figures — only a sparse ~6–16% modeled band
- Press about the person; peer-review invitations unless self-listed
- Patent grants vs. filings, and patent significance
- Recommendation letters, RFE / petition history
- Actual citizenship / visa status — inferred at best, never confirmed
- The "specific endeavor" narrative NIW Prong-1 now demands (Jan 2025)
Defensible product claim: "ranked propensity + intent triage, evidence-tagged, for an attorney to qualify" — never "qualifies for a visa." Plus the parent research's guardrails: flat-fee not revenue-share (ABA 5.4/7.2), OFAC/SDN screen every batch, GDPR/CCPA geo-scope, upstream-provenance legal opinion before scaling, brand vendor-scrub.
Live legal wildcard — track, don't bake in. On 28 Jan 2026 the U.S. District Court of Nebraska (Mukherji v. Miller) held that USCIS's Kazarian step-two "final-merits determination" has no basis in the regulation. Being litigated, not nationwide binding yet. If it holds, clean criteria-counting (3 of 10) becomes more decisive — which would raise the value of a clean professional-profile box-mapping. Treat the gate/eminence balance in §05 as a model assumption with a dial, not a constant. (Also a real talking point for the buyer — the scoring is policy-aware.)