Sixteen years of building first#
I founded Santifer iRepair in Seville in 2009. Over sixteen years the business completed 30,000+ repairs, and along the way I automated everything worth automating: an omnichannel AI agent (Jacobo) ended up resolving around 90% of customer inquiries without human intervention, on top of a business operating system of 12 interconnected databases.
In 2025 I sold the business as a going concern. Without the business for the first time in sixteen years, I faced the question everyone answers with a PDF: finding a job.
The job search as an operated pipeline#
I treated the search the way I would have treated any process in the business: a pipeline with stages, criteria and data. A multi-agent system built with Claude Code scored every listing on a multi-dimensional A-F rubric, generated ATS-optimized PDF résumés tailored to each posting, and pre-filled applications with Playwright. The design principle: automate the analysis, never the decisions. Every application passed through my hands before going out.
740
listings evaluated
68
applications
12
interviews
1
signed offer
The pipeline numbers, frozen as history: 740 job listings evaluated, 68 applications sent, 12 interviews, 1 signed offer. That funnel, 740 down to 1, is the complete story of my 2026 job search.
The full system, with architecture and metrics, is documented in the career-ops case study.
Open-sourcing the system#
When I stopped needing it, I released it under the MIT license: no paywall, no premium tier, free. I said it in Business Insider then and I stand by it: I did not feel comfortable charging people who are looking for work, because finding a job is a basic human need.
It went viral: 12,000+ stars in the first two days. Today career-ops has 60.8K+ GitHub stars (as of July 2026), 12.0K+ forks, 180+ contributors and a Discord community of 4,200+ members. Business Insider (April 2026) and WIRED Greece (April 2026) covered it as a case of AI rebalancing the hiring funnel from the candidate side.
The reversal#
The thesis of the project fits in two sentences: "Companies use AI to filter candidates. I just gave candidates AI to choose companies."
And then the process fully reversed: the CEO who hired me as Head of Applied AI did not find me through an application: he found me through the system I had built and open-sourced. I never applied for that role. Building in public was the résumé.
Out of that experience came The CareerOps Manifesto (July 2026): 9 candidate rights in the AI era, with 38 signatures as of July 2026.
The full manifesto lives at career-ops.org/manifesto.
How it runs today#
I still work full-time as Head of Applied AI. career-ops was built and is maintained evenings and weekends around my full-time job, with a fleet of AI agents doing the mechanical maintenance work: triage, testing, review briefs and releases.
That maintenance system has its own article: agentic maintenance.
Frequently asked questions#
How were you hired without applying?
The CEO of the company found career-ops and the public portfolio documenting it, and reached out directly. There was no application, no résumé sent, no standard process: the system I built to search for a job became the proof of work that landed the Head of Applied AI role. That reversal of the funnel (the employer finds the candidate through their public work) is what the career-ops story calls "the reversal", and it is why the project argues for building in public: in the AI era, your side project is your portfolio.
Are the numbers 740, 68, 12 and 1 real?
Yes, and they are frozen as the historical record of the early-2026 search: 740 job listings evaluated by the system with a multi-dimensional A-F scoring rubric, 68 applications sent after human review, 12 interviews, and 1 signed offer. Every evaluation was logged by the system itself (the run data is in the case study). The open-source project numbers (stars, forks, community) are different: those are live, they grow daily, and on this page they are always shown with their date ("as of July 2026") so no quotation silently goes stale.
Why open source and free instead of a paid product?
Because charging for a job-search tool creates an incentive misalignment: the people who need it most are, by definition, at the moment they can least afford it. career-ops is MIT-licensed, with no paywall and no premium tier; funding is voluntary (GitHub Sponsors). As I told Business Insider in April 2026: I did not feel comfortable charging people who are looking for work, because finding a job is a basic human need. The side effect is that the methodology (automate the analysis, not the decisions, with human oversight) became a shared standard instead of a closed product.
Let's talk
If you have an interesting problem where AI is the right tool, write me.