Career-ops: open-source Claude Code system that filters job applications using engineering discipline instead of spray-and-pray
Career-Ops: The Engineer Who Turned Job Hunting Into a Systems Problem
Most people treat a layoff like a weather event. Something that happens to you, that you wait out. You polish the resume, open LinkedIn, and start clicking Apply on anything that looks plausible. Three months later you’ve sent out 200 applications and heard back from four companies, none of which were your first choice. The process feels random because, for most people, it is random.
One engineer decided to treat it as an engineering problem instead. He got laid off, spent time with Claude Code, and ended up evaluating over 740 job listings through a system he built himself. He landed a Head of Applied AI role. Then he open-sourced the whole thing under the name career-ops. It already has 8,200 GitHub stars.
That’s a good story. But the part worth sitting with is not the automation. It’s the philosophy baked into the architecture.
The Filter Is the Feature
Career-ops has a hard rule: it will not recommend applying to any role scoring below 4.0 out of 5. That’s not a setting you can tweak. That’s the point of the tool.
Most job search automation goes the other direction. It removes friction from applying, which means you apply to more things, which means you spend more time interviewing for roles you never really wanted. Career-ops flips that. It adds friction to applying by raising the bar for what counts as worth your time. The automation is in the evaluation, not the submission.
The system evaluates your CV against each job description through actual reasoning about fit, not keyword overlap. It runs batch evaluations of 10 or more listings in parallel using Claude sub-agents, so the filtering happens fast even when the volume is high. What you get back is a structured A-F score, salary research, ATS-optimized PDF tailored to that specific role, and a tracker entry. One command. One URL. Full output.
The Story Bank is the Part People Will Miss
Buried inside career-ops is something called the interview Story Bank. Every time you run an evaluation, the system accumulates STAR-format answers (Situation, Task, Action, Result) with a reflection layer added. It keeps running until you have somewhere between five and ten master answers that can flex to cover any behavioral question.
That’s a compounding asset. After your twentieth evaluation, you have a library of interview responses refined against twenty different job descriptions. By the time you walk into an interview, you’re not improvising. You’re drawing from a bank you built systematically.
This is what separates career-ops from a fancy wrapper around a job board. The artifact the system produces is not just an application. It’s interview readiness, building up in the background every time you run the tool.
Built to Be Rebuilt
The system ships with 14 skill modes (evaluate, scan, pdf, batch, apply, deep research, negotiation scripts, LinkedIn outreach), a portal scanner pre-loaded with 45 companies including Anthropic, OpenAI, Stripe, and Vercel, and 19 search queries spanning Ashby, Greenhouse, Lever, Wellfound, and Workable. The Go terminal dashboard is built with Bubble Tea.
What’s actually clever is the meta-capability. Because career-ops is built on Claude Code skills, you can ask Claude to rewrite the system itself. “Add these ten companies.” “Change the archetypes to backend roles.” The agent reads the same files it uses to operate, so it knows exactly what to edit. The tool is its own configuration interface.
That’s a design choice, not a gimmick. It means the system adapts to your situation rather than forcing you to fit its assumptions.
Why “Career-Ops” Is the Right Name
Operations implies repeatability, instrumentation, and continuous improvement. It implies you are running a process, not reacting to circumstances. The name is doing real work here.
The engineer who built this did not just automate job applications. He defined what a good opportunity looks like for him, built a system that enforces that definition consistently across hundreds of evaluations, and generated compounding assets along the way. That is operations thinking applied to a personal problem.
Most people will not do this. They will keep clicking Apply on roles they feel lukewarm about, hoping volume covers the gap. For the people who do engage with something like career-ops, the advantage is not just efficiency. It’s that they show up to every interview having already done the reasoning about why this role fits, with story answers sharpened against the actual job description, and a clear floor below which they won’t accept.
The job market is genuinely difficult right now for a lot of people. A tool that helps you stop applying to things that aren’t right for you, and go deeper on the ones that are, is more valuable than a tool that helps you apply faster. That’s a bet worth making.
The repo is at https://github.com/the-momentum/career-ops
🔧
Sources
#AI #JobSearch #CareerOps #ClaudeCode #OpenSource #MLEngineering #AppliedAI
