TalentMatch is an AI-powered recruitment tool that takes a candidate's CV and a job description and instantly returns a structured analysis: a match score out of 100, identified strengths and gaps, a recommendation on whether to apply, and a fully rewritten CV tailored to the role — all without leaving the browser.
Key Features
AI Match Scoring: Gemini 2.5 Flash compares the CV against the job description and returns a structured score (0–100) with color-coded feedback using a circular progress indicator.
Apply Recommendation: The model predicts whether the candidate should apply based on the overall fit, surfaced as a clear Apply / Don't Apply badge.
Strengths & Gaps Analysis: Specific, actionable bullet points highlighting what the CV covers well and what it is missing for the target role.
CV Rewriter: Produces a fully restructured CV — same truthful content, re-emphasized and reordered to target the job — with layout selection (Modern, Classic, Minimal, Tech).
PDF Export: One-click download via jsPDF with proper page-break handling, right-aligned dates, and a clean typographic hierarchy across all sections.
Structured AI Output: Gemini's responseSchema enforces a strict JSON shape, eliminating manual parsing and fence-stripping entirely.
Technologies Used
Next.js 15 (App Router): SSR-ready architecture with dynamic imports and ssr: false for PDF-heavy components.
Google Gemini 2.5 Flash: Structured output via responseMimeType: application/json and a typed responseSchema for zero-parse AI responses.
TypeScript: End-to-end type safety from API response to PDF generation.
jsPDF: Client-side PDF generation with pre-measured block heights to prevent content overflow across pages.
react-circular-progressbar: Score ring with smooth animated fill and color-coded thresholds.
Tailwind CSS: Utility-first styling with custom CSS variables for theming.
Challenges and Learnings
The main challenge was enforcing a predictable structure from a generative model. Using Gemini's responseSchema with typed properties solved this cleanly — no regex, no fence-stripping, no defensive fallbacks needed at the parsing layer.PDF generation required careful handling of multi-line text blocks: pre-measuring total block height before drawing ensured that job entries, education rows, and skill lists never split across page boundaries or overflow into the bottom margin.
TalentMatch is deployed to production and works end-to-end: upload a CV, paste a job description, and receive a full analysis with a downloadable rewritten CV in under ten seconds.
