MyCoverLetters.com

The Problem: Job seekers spend hours writing cover letters that all sound the same. AI generators existed, but they produced generic output because they didn't ask enough questions. A single prompt can't capture the nuances of why someone is a good fit for a specific role.

The Approach: Instead of one-shot generation, I built an iterative refinement flow. The user uploads a resume and pastes a job posting. Claude generates a first draft. Then the system asks targeted follow-up questions — "The job mentions Kubernetes experience. Your resume shows Docker but not K8s. Should we highlight your containerization experience or skip this?" Each round produces a meaningfully better letter.

What Was Hard: The billing model was more complex than the AI integration. Four pricing tiers (free trial, per-letter credits, monthly subscription, annual), plus a referral program with leaderboard tracking. Stripe webhook handling for subscription changes, failed payments, and credit top-ups required careful state management. I also implemented sodium encryption for professional references — users trust the platform with names and phone numbers of people who vouched for them, so that data gets encrypted at rest.

The Result: Full SaaS stack — Stripe subscriptions, PDF/DOCX export via DomPDF and PHPWord, referral tracking, and ARIA CI/CD. The iterative approach produces dramatically better output than one-shot generators. Live at mycoverletters.com.