Strawberry Matcha
AI agent for marriage-based green card applicants (CR1 or F2A) filing without a lawyer.
- Role
- AI UX Designer · Solo project
- Tools
- Cursor, Claude API, Supabase, Figma
- Focus
- Conversational AI, Decision-support UX
- Timeline
- 3 weeks · v0 → v1
Context
An AI agent for couples filing their marriage-based green card alone.
Strawberry Matcha is for people going through the CR1 or F2A green card process without a lawyer. The applicant talks to it the way they would talk to someone who already knows the process. As they chat, it asks follow-up questions, picks up details about the case, and gets more useful the more it's used.
User problem
Filing alone leads to mistakes. General AI makes it worse.
Most couples filing for a marriage-based green card don't have a lawyer. They piece the process together from USCIS pages, Reddit threads, and lawyer blogs from years ago. Three forces collide.
Lawyers are unaffordable for many couples.
Marriage-based green card lawyers charge $2,000 to $8,000+ on top of $1,700+ in USCIS fees. For most couples that's not a real option, so they file alone and learn the process as they go.
DIY filing produces costly mistakes.
1 in 4 marriage-based applicants gets a Request for Evidence, most of them for missing documents and filing errors. The kind of mistakes that happen without someone to ask the right questions first.
Source: CitizenPath, USCIS Request for Evidence Guide
General AI hallucinates on legal workflows.
ChatGPT and Claude sound confident, but they aren't trained on legal workflows. They miss steps. They give advice that fits a generic case, not yours. For an immigration filing, that's not safe enough.
Problem statement
How might we make case-aware legal guidance affordable and reliable for couples filing alone?
Solution
A legal-trained AI agent that learns your case through conversation.
Three pillars hold it together. Each one is a deliberate counter to the way filing alone usually breaks down.
Built around how immigration lawyers actually work.
Strawberry Matcha follows the legal workflow, asks the right questions in the right order, and stays grounded in the applicant's specific case instead of falling back on generic advice.
It learns as you chat.
Instead of a long form upfront, it starts with a short intake and keeps learning as the user talks. It asks follow-ups when it needs more context, and suggests what to ask when the user isn't sure.
It walks you through every form, field by field.
When the user uploads a USCIS form, Strawberry Matcha checks each field against the case it has built. It points out what's missing, what looks off, and how to fix it before submission.
Key features
Three features that close the gap between filing alone and having a lawyer.
Ask Strawberry Matcha, a conversation that knows your case.
Users can ask anything, anytime. Strawberry Matcha answers based on the applicant's actual case status and preparation progress, and updates the case as the conversation continues.
Field Translator, fills the gap between your real life and the form.
The applicant uploads a USCIS form PDF (any edition). The system reads the actual fields, cross-references the case data, and tells the user what to enter in each one, including the trickiest part: format conversion. A Korean address gets reshaped into US form structure. A Korean name gets matched to passport romanization. The values come out form-ready.
User input
Free-form Korean address as the applicant naturally writes it.
USCIS form output
- ProvinceSeoul
- City or TownGangnam-gu
- Street NameTeheran-ro
- Street Number123
- Apt / UnitDong 101, Ho 202
Parsed and reformatted into the exact fields each USCIS form expects.
Timeline guidance, so you know where you are and what's next.
Each milestone shows where the applicant is in the process, what the step actually means, and what usually happens next, so the case never feels like a black box.
Design process
From concept to crafted product in five steps, plus two design decisions that defined the shape of it.
- 01
Define concept & Research to train the AI
Mapped how immigration lawyers actually walk a couple through CR1 / F2A.
- 02
Design System Architecture
Used Cursor's plan mode to map out the full system as a diagram, so I could see how every piece fit before writing code.
- 03
Fast validation
Used Cursor to spin up a working prototype quickly, so I could test the idea with real applicants before investing more.
- 04
Iterations
Reworked chat structure and onboarding based on where trust was breaking.
- 05
Craft refinement
Polished the UI in Figma, tightening tone, pacing, and visual hierarchy across the whole product.
Reducing cognitive overload in chat.
The first version of the chat dumped each response into one long paragraph. The AI was answering well, but the answers weren't usable. People scanned, hesitated, and stopped.
I explored three ways to give responses more shape before settling on one.
Single response paragraph.
- Fast to implement; no extra UI.
- Buries what matters most.
- Users don't know what to ask next.
Full doc-style hierarchy with headers and bullets.
- Maximum scannability.
- Loses conversational warmth.
- Overkill for short answers.
Two-layer voice (serif acknowledgment + sans-serif info) with suggested follow-ups.
- Reads warm and human.
- Scannable at a glance.
- Nudges the next question.
- More design and prompt work.
Final pick:
Option 03
Two-layer voice keeps the chat warm but makes the answer scannable, and the suggested follow-ups stop users from getting stuck on what to ask next.
Onboarding as the foundation of trust.
The original onboarding was too short. Without enough context about the user's case, the AI was filling gaps by guessing, and the hallucinations broke trust fast.
I studied how immigration lawyers actually intake their clients. The questions they ask up front aren't paperwork. They're how the lawyer learns the case before giving any advice. I rebuilt onboarding around those same questions, so the AI starts with enough context to be accurate from the first message.
Reflection
What I took away from designing an AI agent for a high-stakes legal workflow.
Designing an AI agent is designing how it thinks.
Most of the work happened underneath the screens. Prompts, follow-up logic, what the AI asks versus what it answers, what it stores about the user. The visible UI was the smallest part.
Onboarding is data acquisition, not a signup.
How well an AI agent performs depends on what it knows going in. Designing onboarding well is designing the AI's first impression of the user, and everything downstream flows from there.
Conversational UX is about pacing, not just tone.
Users filing alone don't need more information. They need information at the right moment, in a shape they can act on, with a clear next step. That's a design problem, not a content problem.