75% of Resumes Never Reach a Human

Overview

AI Resume Tailor is a bilingual SaaS application that helps job seekers understand exactly why their resumes get rejected by Applicant Tracking Systems and what to do about it. Users paste their resume and a job description, and the system returns a structured analysis with a match score, missing keywords, and severity-coded suggestions.

The application runs a React 19 frontend deployed on Vercel with a Flask backend on Render. The AI engine uses OpenAI GPT-4 with structured JSON output and ATS-focused prompts to deliver deterministic, actionable feedback. Authentication is handled by Clerk (Google, LinkedIn, email/password), payments by Stripe with embedded checkout, and user data lives in Neon serverless Postgres.

The product is designed around zero friction — guests get 5 free analyses with no signup required. Privacy is enforced by architecture: no resume data is stored, and all processing happens in real time.

The Challenge

• ATS Black Box: Job seekers submit resumes into systems they can't see. Most have no way to know whether their resume will pass keyword filters.
• Generic Advice: Existing resume tools provide surface-level feedback rather than job-specific, keyword-level analysis tied to the actual posting.
• Friction Barriers: Competing tools require account creation or payment before delivering any value.
• Language Gap: The Latin American job market is underserved by English-only tools.

The Solution

Structured ATS Analysis: GPT-4 returns structured JSON with a match score, keyword gaps, and severity-coded suggestions — all parsed into type-safe TypeScript interfaces on the frontend.

Zero-Friction Access: Guest users get 5 free analyses with no account required. The product delivers value before asking for anything in return.

Bilingual by Design: Full i18n coverage across every UI element, error message, and analysis output. English and Spanish are first-class citizens.

Privacy-First Architecture: No resume data is persisted. Analysis happens in a single request/response cycle.

Technical Highlights

• Structured JSON Output: GPT-4 with strict JSON mode and low temperature (0.1) for deterministic responses
• Type-Safe API Contract: TypeScript interfaces with union types and transformation functions for backward compatibility
• Modular Analysis Engine: Separate scoring, keyword extraction, and suggestion modules — each independently testable
• Severity-Coded UI: Red/amber/blue visual hierarchy maps directly to AI output severity levels
• Embedded Stripe Checkout: Guest-to-paid conversion with post-payment account creation
• Clerk Webhook Pipeline: Automatic profile provisioning in Neon Postgres across all auth methods
• Skeleton Loading States: Layout-perfect skeletons matching final render dimensions — zero layout shift

Results

For the End User:

• Structured, actionable ATS feedback in under 10 seconds
• 5 free analyses with no signup, no email, no commitment
• Native Spanish support — not machine-translated afterthoughts
• Zero data retention — resume text never persisted

Technical Demonstration:

• Full-stack production SaaS with AI, auth, payments, and database
• Clean separation between AI engine, scoring logic, and presentation layer
• Internationalization architecture that scales to additional languages without refactoring
• Privacy-by-design approach that eliminates an entire category of data liability

This project demonstrates end-to-end product engineering: from AI prompt design through payment integration to bilingual UX — shipping a production SaaS that solves a real problem for a real audience.