Settle

Inspiration

Millions of immigrants arrive in the US each year with rich financial histories — CIBIL scores, tandas, UPI, paluwagan, esusu — only to be told their entire financial identity doesn't exist here. They're not starting from zero knowledge; they're starting from zero recognition. We built Settle because the gap isn't financial literacy — it's translation. If you've run a chit fund, you already understand rotating savings. If you've used UPI, you already get instant transfers. You just need someone to show you where those concepts live in the American system.

What it does

Settle is a personalized financial onboarding guide for immigrants. Users complete a short 7-question flow covering their home country, immigration status, SSN situation, financial tools they used back home, and their most urgent financial concern. Settle then maps their home financial tools (CIBIL scores, tandas, paluwagan, esusu, GCash, etc.) to their direct US equivalents, explains the key differences, and surfaces a prioritized action roadmap. A full first-year checklist covers banking, credit, taxes, remittances, and savings — with items that unlock a growing city skyline as users complete them.

How we built it

The stack is Next.js 14 with the App Router, Tailwind CSS, and TypeScript. The knowledge graph lives in a structured JSON file mapping 23 home financial concepts across Mexico, India, the Philippines, Nigeria, and Central America to their US equivalents — with similarity scores, urgency rankings, and caution flags. An AWS Bedrock (Claude Sonnet) integration generates personalized portrait text and action recommendations based on the user's profile. The graph queries run through AWS Neptune with a local fallback so the app works even without a cloud connection. The onboarding globe is built with react-globe.gl and Three.js, with custom point highlighting for the supported countries.

Challenges we ran into

Getting the 3D globe to center properly across different screen sizes and parent container layouts required fighting through react-globe.gl's internal canvas sizing — the fix was passing explicit width and height props directly to the Globe component rather than relying on CSS. Prompt engineering for Bedrock was tricky: we needed Claude to stay strictly within the verified knowledge base without hallucinating financial products or regulations, which required a carefully structured system prompt with explicit grounding instructions. Handling the async handoff between the loading page and result page — ensuring the AI response was ready before navigation while still showing a minimum animation duration — required coordinating Promise.all with both a timer and the API fetch.

Accomplishments that we're proud of

The knowledge graph concept-to-concept mapping is genuinely useful — telling someone that their chit fund experience translates directly to a lending circle that also builds their FICO score is the kind of insight that changes behavior. The skyline gamification mechanic, where completing financial milestones literally builds a city, captures the "building your new life" metaphor in a way that feels earned rather than gimmicky. The app degrades gracefully — if Bedrock is unavailable, the knowledge graph still delivers value.

What we learned

Financial systems are far more culturally embedded than they appear. The tanda/chit fund/paluwagan/esusu are essentially the same product operating across four continents — but the US equivalent (lending circle through a CDFI) adds a credit-building dimension that fundamentally changes its value. We also learned that immigrants don't need more information — they need better translation. The framing "you already know how this works, here's what's different" consistently tested better than starting from scratch.

What's next for Settle

Expanding the knowledge graph to cover more countries and more nuanced products — BNPL equivalents, microfinance, cooperative banking. Adding a conversational layer so users can ask follow-up questions about specific concepts. Partnering with credit unions and CDFIs to surface real, vetted product recommendations rather than generic guidance. Building a community layer where immigrants who've navigated the system can add notes and tips to specific concept cards. And eventually, a mobile app with offline support — because many of the people who need this most have unreliable internet access.