Problem
I kept finding the same jobs across different sources, manually reviewing descriptions and trying to keep resumes, facts and projects synchronized. Repetitive work consumed time before the application decision itself.
PERSONAL PRODUCT · WINDOWS · LOCAL-FIRST
PERSONAL PRODUCT IN USECarreiraPessoal
Job discovery, evidence, matching and resume routing
I built this product to reduce repetitive work in my own job search without handing the final decision to an AI: it organizes sources, removes duplicates, checks career direction, protects evidence and helps select the right resume.
Validation283 passing Python tests
Coverage102 ATS/platform families · 11 direct collectors
StackFastAPI · React/TypeScript · Tauri/Rust · SQLite/FTS5
What I built
Multi-source discovery, normalization, deduplication, URL/text/file input, Career Goal, a versioned professional profile, EvidenceGuard, Resume Router/Fitness, an application timeline and a browser assistant.
AI as reinforcement
The core works without an AI provider. Local matching, career rules, deduplication and safeguards remain available; Ollama, LM Studio and external providers are optional capabilities that can be disabled.
Human boundary
The system can help prepare an application, but it does not invent experience and does not submit on its own. CAPTCHA, site terms and final submission remain human; sensitive claims need sufficient evidence.
Architecture
Sources / URL / file → normalization → deduplication → Career Goal → matching → EvidenceGuard → Resume Router → application → browser assistant → manual submission.
Privacy
The public repository is a sanitized edition. Personal databases, real applications, credentials and third-party data are not published.
The public evidence prioritizes architecture, rules and sanitized code. The complete v12.5.2 build is the personal product I use; its test count validates product contracts, not universal AI accuracy.
Next projectOperational Procurement Catalog
Contact