CVBranch — Evidence-Based Resume Tailor
An AI resume tailor for software and IT roles. Upload a CV, paste a job description, and CVBranch shows required vs preferred skill gaps, asks for real experience, then composes a one-page resume for that application. Matching is deterministic; LLMs draft only from facts you confirm — no invented stack, metrics, or seniority. I designed and built CVBranch end-to-end: CV and JD parsing, a deterministic matcher, evidence interviews, a one-page resume composer, and the product UI. I also designed the LLM layer — structured prompts, grounding and anti-hallucination rules, and post-processing that strips ChatGPT-style phrasing — using OpenAI (GPT-4.1 mini) in production with a local Ollama path for development. The stack is a TypeScript monorepo with Next.js on the frontend and Firebase Auth, Cloud Functions, Firestore, and Storage on the backend.






