DebtClear

The AI debt strategist that tells you exactly which debt to kill first, writes your negotiation script, and simulates your path to zero.

Comment

Inspiration

Debt in America is a $17 trillion crisis — but the tools people use to fight it are spreadsheets from 2009. I watched people in my life paying minimums for years, not because they lacked discipline, but because no one ever showed them the math. The difference between Avalanche and Snowball strategies can be worth $8,000 and 3 years of your life. That gap shouldn't exist.

I built DebtClear to be the financial advisor everyone deserves but most people can't afford.

What it does

DebtClear is a full-stack AI debt strategy engine. You enter your debts — balances, APRs, minimums — and in seconds you get:

  • Month-by-month simulation of Avalanche vs. Snowball payoff strategies with exact dollar and time savings
  • A Financial Stress Score (0–100) that quantifies your debt load relative to income
  • AI advisor chat powered by Llama 3.3 70B — ask "what if I lose my job?" or "should I refinance?" and get answers grounded in your actual numbers
  • Negotiation scripts — word-for-word phone scripts to settle each debt for less, with hardship framing, opening offers, and counter-offer responses
  • Phone roleplay — practice the call against an AI collections agent (Sarah) who pushes back and only settles if you make a strong case
  • Settlement letter generator — formal certified-mail letters ready to print and send
  • Milestone timeline — visualize every debt payoff date on a single chart
  • Downloadable PDF plan — a full editorial-quality document with your strategy, debt ledger, and action steps

How I built it

Backend: Django REST API running on AWS EC2. A deterministic month-by-month debt simulator handles both Avalanche (highest APR first) and Snowball (lowest balance first) strategies. The stress score is calculated from debt-to-income ratio, weighted average APR, and minimum payment burden. All AI calls route through Groq's API with a multi-key pool and automatic failover for reliability under rate limits.

Frontend: Next.js 14 App Router with TypeScript, Tailwind CSS, and Framer Motion. The UI is split across four routes (/, /how-it-works, /analyze, /results) with a persistent WebGL nebula shader as the background. Charts use Recharts. PDF export uses jsPDF. The phone roleplay uses the browser's Web Speech API for voice input and speech synthesis.

AI: Llama 3.3 70B via Groq for all generative features (analysis, negotiation scripts, advisor chat, roleplay, settlement letters). Every prompt is grounded in the user's exact financial snapshot — no hallucinated numbers.

Challenges I ran into

  • Groq rate limits at scale — solved with a multi-key pool that round-robins and fails over automatically, with hard timeouts so the UI never hangs
  • WebGL shader on mobile — the nebula fragment shader was killing mobile frame rates. Fixed with adaptive resolution scaling and a 30fps cap on mobile
  • jsPDF font rendering — Unicode arrows and special characters silently render as garbage in built-in fonts. Learned to stick to ASCII-safe characters for all PDF content
  • Debt simulation accuracy — minimum payments recalculate monthly as balances fall. Getting the compounding math right (especially the waterfall when a debt is paid off) required careful month-by-month state tracking

What I learned

  • How to build a real financial simulator that handles edge cases (zero-balance debts, minimum-only scenarios, payment waterfalls)
  • How to ground AI outputs in user data so they're actually useful rather than generic
  • Web Speech API quirks across browsers and how to gracefully fall back to text input
  • The difference between a feature that looks good and one that actually helps someone — the negotiation roleplay went through several iterations before it felt like genuine practice, not a gimmick

What's next

  • Plaid integration for automatic debt import
  • Notification system to remind users of their next attack target
  • Creditor-specific negotiation intelligence (different scripts for Capital One vs. medical debt vs. student loans)
  • Multi-currency support

Built With

Share this project:

Updates