When I was searching for a job right out of college, I had a spreadsheet to track all my applications. While that is a workable method, one developer has leapfrogged it with an AI pipeline instead. Santiago, a former founder and now Head of Applied AI, spent his job search engineering his way out of the search problem. Manually sifting through listings and tailoring CVs one by one is a pretty time consuming task so he built Career-Ops, an AI-powered job search system on top of Claude Code. He used it to evaluate over 740 job offers, generate more than 100 tailored CVs, and ultimately land his current role. The best thing is that he has now open-sourced the entire thing.
Also read: Microsoft Copilot terms are legal ‘LOL’ for every business using it
Career-Ops runs off a single slash command inside Claude Code. Paste in a job URL or description, and it takes over – scraping the listing, scoring the opportunity, generating a tailored CV, and logging everything to a tracker automatically.
The scoring engine evaluates each role across ten weighted dimensions, assigning a grade from A to F. The system is deliberately anti-spray-and-pray. Applying to fewer, better-fit roles beats blasting your CV everywhere any day of the week and the scoring system forces that discipline.
Also read: Claude AI has functional emotions that influence behaviour, Anthropic study finds
The repo ships with 14 skill modes covering the full job search lifecycle. A portal scanner comes pre-configured with 45+ companies – from Anthropic and OpenAI to ElevenLabs and n8n – searching across Ashby, Greenhouse, Lever, and Wellfound simultaneously. When it finds a match, it generates an ATS-optimized PDF using Playwright, with keyword injection and custom typography baked in.
There is also a batch mode for processing ten or more offers in parallel using Claude sub-agents, a form-filling mode for live applications, and a LinkedIn outreach generator. A terminal dashboard built in Go lets you browse, filter, and sort your entire pipeline without leaving the command line.
What makes Career-Ops genuinely interesting is not just what it automates, but how it is designed. The system reads and edits its own configuration files, meaning you can ask Claude to change scoring weights, swap archetypes, or add companies to the scanner and it does it. The AI is not a feature bolted on. It is the interface. The repo is live on GitHub and free to use.
Also read: Human context missing: AI benchmarks are flawed, researcher explains why