CVCrew

Problem

Job hunting is scattered and repetitive: tailoring the CV for every listing, tracking applications, writing cover letters — each needs a separate tool and separate effort. A single AI chat can't carry the depth of that process; context gets lost and output turns shallow.

Solution

CVCrew is a career copilot where multiple AI agents work together. Each agent has a single responsibility — CV analysis, listing matching, content generation — and an orchestration layer coordinates them. Task delegation between agents is built on Claude and the Anthropic SDK; the user watches agent progress in real time over WebSocket.

What I learned

• Multi-agent orchestration: splitting work into specialized agents instead of one giant prompt improves both output quality and debuggability.
• Streaming UX: showing live progress during long AI jobs is the main factor in perceived speed.
• Prompt engineering: defining agent roles with sharp boundaries keeps agents from stepping on each other's work.