I built StoicCredit after repeatedly running into credit-related bottlenecks while working on small projects (micro-SaaS, services, short-term rentals).
The core idea is simple: instead of generic credit advice, take a real credit report and turn it into a concrete, time-sequenced action plan based on what’s actually in the file.
How it works at a high level:
- Users upload credit report PDFs (Experian, TransUnion, Equifax)
- The system extracts and normalizes accounts, limits, utilization, payment history, inquiries, and derogatories
- A rules-based layer generates a prioritized roadmap (what to fix first, what to wait on, and when applying makes sense)
Some technical details that might be interesting:
- PDF text extraction + LLM-assisted structured parsing (used for normalization, not scoring)
- A deterministic rules engine layered on top of parsed data (no black-box “AI score”)
- React Flow for visualizing credit-building paths and dependencies
- A small internal dataset mapping lender bureau pulls and common underwriting constraints
Stack:
- Frontend: React
- Backend: Firebase Cloud Functions
- LLM: Claude (used for parsing and explanation, not decision-making)
- Auth/Billing: Firebase + Stripe
There’s a free tier to try it out.
Happy to answer questions about:
- Credit report PDF parsing edge cases
- How I validated rules without over-promising outcomes
- Where LLMs helped vs. where they hurt reliability
I built StoicCredit after repeatedly running into credit-related bottlenecks while working on small projects (micro-SaaS, services, short-term rentals).
The core idea is simple: instead of generic credit advice, take a real credit report and turn it into a concrete, time-sequenced action plan based on what’s actually in the file.
How it works at a high level: - Users upload credit report PDFs (Experian, TransUnion, Equifax) - The system extracts and normalizes accounts, limits, utilization, payment history, inquiries, and derogatories - A rules-based layer generates a prioritized roadmap (what to fix first, what to wait on, and when applying makes sense)
Some technical details that might be interesting: - PDF text extraction + LLM-assisted structured parsing (used for normalization, not scoring) - A deterministic rules engine layered on top of parsed data (no black-box “AI score”) - React Flow for visualizing credit-building paths and dependencies - A small internal dataset mapping lender bureau pulls and common underwriting constraints
Stack: - Frontend: React - Backend: Firebase Cloud Functions - LLM: Claude (used for parsing and explanation, not decision-making) - Auth/Billing: Firebase + Stripe
There’s a free tier to try it out.
Happy to answer questions about: - Credit report PDF parsing edge cases - How I validated rules without over-promising outcomes - Where LLMs helped vs. where they hurt reliability