Laid-Off Engineer Open-Sources AI Job Search System 'career-ops'

An engineer who was recently laid off has built and open-sourced a comprehensive AI-powered job search system called career-ops. The system, constructed using Claude Code, helped him evaluate over 740 job offers and ultimately secure a Head of Applied AI position. The project has rapidly gained traction on GitHub, amassing 8.2k stars under an MIT license.

What career-ops Does

career-ops is a terminal-based pipeline designed to replace the manual, spreadsheet-heavy job search process. At its core, it's a filter, not a spam tool. The system ingests a job URL and returns a structured evaluation, an ATS-optimized PDF resume tailored to that specific role, salary research, interview preparation notes, and logs the entry into a tracking system—all from a single slash command.

The tool operates on a philosophy of quality over quantity. It explicitly refuses to recommend applying to any job that scores below 4.0 out of 5.0 in its evaluation. The evaluation is based on reasoning about the fit between a user's CV and the job description, moving beyond simple keyword matching.

Technical Architecture & Features

The system is built as a collection of Claude Code skills, making it inherently modifiable by instructing Claude itself to rewrite parts of the system. The codebase includes:

• 14 Skill Modes: Including evaluate, scan, pdf, batch, apply, deep_research, negotiation_scripts, and linkedin_outreach.
• Portal Scanner: Pre-loaded with over 45 company career portals, including Anthropic, OpenAI, ElevenLabs, Mistral, Cohere, Stripe, Retool, Vercel, and Decagon.
• Multi-Platform Search: 19 pre-configured search queries across job platforms like Ashby, Greenhouse, Lever, Wellfound, and Workable.
• ATS-Optimized PDF Generation: Uses Playwright for PDF rendering with specific fonts (Space Grotesk, DM Sans) known for high ATS parsing success.
• Go Terminal Dashboard: A pipeline browser built with the Bubble Tea TUI framework.
• Batch Processing: Can evaluate 10+ job offers in parallel using Claude sub-agents.
• Interview Story Bank: Automatically accumulates STAR (Situation, Task, Action, Result) + Reflection narratives across all job evaluations, building a repository of 5-10 master answers for common behavioral questions.
• Application Auto-fill: Can populate common application form fields.

How It Works: The Pipeline

1. Input: A user provides a job description URL or uses the built-in search to find roles.
2. Evaluation & Reasoning: The system's core evaluate skill uses Claude to analyze the job description against the user's CV. It doesn't just match keywords; it reasons about experience alignment, skill gaps, and cultural fit, outputting a letter grade (A-F) and a score out of 5.
3. Material Generation: For jobs that pass the filter (≥4.0), the pdf skill generates a tailored resume. It reformats the user's base CV to highlight the most relevant experiences for the specific role, using ATS-friendly formatting.
4. Research & Prep: Concurrently, it runs deep_research on the company and role, compiling salary bands and potential interview questions.
5. Tracking & Continuity: All data is logged. Stories developed during evaluation are saved to the central Story Bank for future interview prep.

Open Source & Modifiability

The entire system is open-source. Its unique architecture as a Claude Code project means users can directly instruct Claude to modify the system. For example, a user can prompt: "Change the evaluation archetypes to focus on backend engineering roles" or "Add these 10 new companies to the portal scanner." Because Claude has access to the same code files it operates on, it can implement these changes directly.

gentic.news Analysis

This project exemplifies a significant trend in the AI tooling landscape: the rise of highly specific, agentic workflows built on top of foundation model APIs. career-ops isn't a general-purpose chatbot; it's a finely-tuned pipeline that orchestrates multiple calls to Claude (for reasoning, generation, summarization) alongside traditional software components (Playwright for PDFs, Bubble Tea for UI). This mirrors the architectural pattern seen in other successful open-source projects like OpenInterpreter or Smol Agents, which also wrap LLMs in deterministic workflows for complex tasks.

Its viral reception (8.2k stars in a short period) highlights a acute market need. The tech job market, particularly in AI, remains volatile. Tools that provide a structured, strategic advantage in job searching—moving beyond LinkedIn Easy Apply—resonate deeply. The choice to build on Claude Code is strategically interesting. It positions the tool within Anthropic's emerging ecosystem and leverages the model's strong reasoning and instruction-following capabilities for the core evaluation task, which is more nuanced than simple retrieval.

Furthermore, the project's philosophy—acting as a filter rather than a spam tool—is a direct critique of the low-effort, high-volume application strategies that have degraded job markets. By enforcing a quality threshold, it attempts to increase the signal for both candidates and recruiters. This aligns with a broader movement towards precision over scale in AI applications, a trend we've noted in areas like AI-powered sales outreach and marketing.

Frequently Asked Questions

Is career-ops free to use?

Yes, the career-ops software is 100% open-source and released under the permissive MIT license. This means you can use, modify, and distribute it freely. However, to run it, you will need your own Anthropic API key to make calls to Claude, and you will incur standard API usage costs based on your usage volume.

How does the AI evaluation work? Is it just keyword matching?

No, according to the developer, the system uses Claude's reasoning capabilities to evaluate the fit between your CV and the job description. It goes beyond keyword matching to assess experience alignment, skill transferability, and potential gaps. It outputs a reasoned analysis along with a score (A-F, 1-5).

What if I'm not a software engineer? Can I still use this tool?

The tool is run from the terminal and requires some technical setup (installing Go, Node.js for Playwright, configuring API keys). While the developer has made it modifiable via natural language thanks to Claude Code, initial setup and operation require comfort with a command-line interface. It is currently most accessible to developers and technically proficient users.

How does the ATS-optimized PDF generation work?

The system uses the Playwright browser automation library to generate a PDF from an HTML resume template. The template uses specific, ATS-friendly fonts (Space Grotesk, DM Sans) and a clean, parsable layout to maximize the chances that Applicant Tracking Systems correctly read and categorize the resume content. It dynamically populates this template with content tailored for each specific job.