CVDesigner
A four-agent platform that rewrites CVs for every job posting.
An end-to-end agentic platform: analyze a job posting, research the target company, rewrite the candidate's CV for that exact role, and score the fit — all grounded in real candidate data, never hallucinated.
- Python
- Agno
- Pydantic
- Gemini API
- pgvector
- PostgreSQL
- Next.js
- Gemma 4
- Qwen 3.6
The problem
Generic CV builders produce slop. Manual tailoring takes hours per application. The ATS drops anything off-keyword. Candidates lose the role before a human ever reads their story.
Four agents, one outcome
job-analysis parses the posting into weighted requirements. research pulls company context from the live web. generation rewrites each CV section with structured output against the weighted brief. scoring evaluates the final document against custom criteria and emits a fit percentage, strengths, and gaps.
Grounded in pgvector
Candidate profiles, prior CVs, and job descriptions are chunked and embedded into pgvector. Every generation call retrieves the most relevant spans of real history first — no agent writes a bullet it cannot cite.
Function calling, shared context
Agents communicate through typed tool calls and a shared context window. When research finds a new keyword, generation picks it up the next turn. When scoring flags a weakness, research revisits the web for mitigating context.
Output you can defend
The final artifact ships with quantitative indicators: alignment percentage, top three strengths to emphasize, top three gaps to address. Every bullet is traceable to a source in the candidate's real history.