Lumova — winning EB1A green cards without $40k in attorney fees

Forty-seven percent of EB1A green-card petitions get denied. Most of those denials aren’t about whether the applicant is actually extraordinary. They’re about how the evidence is framed.

USCIS officers decide petitions by measuring evidence against the regulatory language in the decisions they’ve already made. Immigration attorneys bill $250–$500 an hour to translate a researcher’s patents, a founder’s press coverage, or an athlete’s championships into that language. For a typical self-petitioner, the total bill lands somewhere between $10,000 and $40,000— money most people don’t have, or shouldn’t be spending.

The outcome is predictable: brilliant people with plainly qualifying track records get rejected because their evidence wasn’t packaged in the right words. Others give up before ever filing. The platform that would fix this didn’t exist — partly because building it is genuinely hard, and partly because the category is loud with generic LLM wrappers that hallucinate regulatory citations and don’t know what winning actually looks like.

Lumova’s founders had watched this happen in their own networks one too many times. They wanted the regulatory expertise of a senior attorney available in an afternoon, for the price of a streaming subscription. That’s the brief they brought to us.

We started with one question: what would it actually take to outperform an unspecialised attorney on petition language, at roughly 2% of the cost?

The answer wasn’t a generic LLM wrapper. Those exist in this category, and they fail at it — they hallucinate case citations, miss criterion-specific framing, and ignore the fact that officers read evidence against the pattern of rulings they’ve already signed off on. We’d have been embarrassed to ship one.

The answer was a curated, version-controlled dataset of over 5,000 published Administrative Appeals Office (AAO) decisions, retrieval-augmented generation (RAG) grounded in that dataset, and prompts engineered per EB1A criterion — so the system can see exactly how an officer wrote about, say, “original contributions of major significance” in a dozen approved entrepreneurship petitions before drafting the applicant’s.

Commercially, we built Lumova as a four-tier SaaS: a free eligibility quiz to qualify intent, a Starter tier for assessment, a Pro tier for drafting, and an Elite tier for full petition assembly with unlimited audits. Stripe-billed, backed by a 90-day money-back guarantee, priced 60–80% below attorney models at every step.

Assessment & evidence roadmap

• A 7-question eligibility quiz that returns an instant readiness score across all 10 EB1A criteria — without asking for a credit card, because qualified intent is worth more than a locked paywall.
• A personalised evidence roadmap that ranks the applicant’s strongest qualifying criteria first, so no one spends three months chasing the wrong angle.
• A pre-filing petition audit with severity-rated action items, so the user knows exactly what needs fixing before an officer ever sees the packet.

AI writing & generation

• Criterion-by-criterion petition sections drafted in USCIS-compliant regulatory language, grounded in retrieval against the AAO dataset.
• Reference letter drafting with the framing, tone, and corroborating detail officers expect — not generic recommendation templates.
• Request-for-Evidence (RFE) response assistance that analyses the specific gaps an officer has flagged and drafts direct, cited rebuttals.

AAO Intelligence

• A searchable interface over thousands of published AAO decisions, with approval-pattern visibility broken down by discipline — so a tech founder sees what actually worked for other tech founders, not academics.
• A real-time USCIS news pipeline pulling from official government sources, so guidance is never quietly out of date.

Commerce

• Four-tier subscription billing (Free / Starter / Pro / Elite), monthly and annual, with upgrade/downgrade paths and a 90-day money-back guarantee.
• Stripe-native billing, dunning, and subscription lifecycle — shipped so the Lumova team could run the business without engineering in the loop.

• Claude Opus via the Anthropic API, with criterion-specific prompt engineering and temperature tuning per output type (petition section vs. reference letter vs. RFE response).
• Retrieval-augmented generation against a curated, versioned vector store of AAO decisions, re-ingested as USCIS publishes new rulings.
• Document export engine producing petition-ready DOCX and PDF outputs with the formatting officers are used to seeing.
• Real-time USCIS.gov sync so the product reflects the most current official guidance, not a snapshot from six months ago.
• Stripe subscriptions for the four-tier plan with monthly and annual options, trial handling, and the money-back guarantee workflow.

These are figures Lumova publishes on their own site. We’re including them here because they’re a matter of public record — not as an independently-audited PeopleAMP claim.

Lumova also publishes a $15,000 average attorney-fee saving per user, and a 6-month median filing timeline compared with the 12+ months typical of traditional attorney-led petitions.

Lumova isn’t a chatbot. It’s a product that has to be right where it matters most — officers reject petitions for the language they’re written in, so that language had to be drawn from actual winning decisions, not invented by a model. Shipping that took more than glue code on top of an API.

It’s also a real commercial product with a real business model: paying subscribers, dunning, refunds, annual/monthly plans, a money-back guarantee that the team can actually honour because the product consistently delivers. Not a side-project demo dressed up for a pitch deck.

If we can ship this — regulated output language, a curated legal corpus, a four-tier Stripe SaaS, paying customers in 80+ countries — we can ship yours. That’s the whole point of putting it here.