hiring-agent

hiring-agent is a developer-oriented resume scoring agent for technical recruiting teams. The repository describes a resume-to-score pipeline that extracts structured data from PDFs, enriches it with GitHub signals, and returns a fair, explainable evaluation. It is not a generic applicant tracking system. It is closer to a transparent scoring engine that teams can run, inspect, and adapt. The workflow starts with a resume PDF. The project parses the file to Markdown, calls an LLM per resume section through Jinja templates, normalizes loose JSON into a JSON Resume style schema, and then looks for GitHub profile information. When a GitHub username is present, the agent fetches profile and repository data, classifies projects, and asks the LLM to select meaningful contributions. The final evaluator prints a report and can append key fields to a CSV file when development mode is enabled. The tool is useful because hiring AI is risky when it acts like a black box. hiring-agent exposes the pipeline pieces: PDF extraction, LLM provider selection, prompts, schemas, GitHub enrichment, category scores, and evidence. Recruiters and engineering managers can inspect why a score was produced instead of accepting a hidden ranking. The README also supports local execution with Ollama, which helps teams test candidate data workflows without sending every resume to a hosted model. Pricing is not SaaS pricing. The source is public under the repository license, while runtime cost depends on the LLM backend you choose. A local Ollama setup can avoid hosted token costs, while Gemini requires a configured API key and any provider usage charges. Teams should test the pipeline on synthetic or consenting sample resumes first, because resume evaluation touches sensitive data and fairness requirements. The strongest fit is internal evaluation support, not fully automated rejection. A hiring team can use the output to organize evidence, flag projects worth reviewing, and make interviews more consistent. Humans should still own the decision, especially when a resume lacks a GitHub profile, uses nontraditional experience, or belongs to a candidate from a background the scoring criteria may not cover well. Builders evaluating hiring-agent should inspect the prompts and scoring templates before use. The value of an explainable pipeline depends on the quality of the rubric and the evidence behind each score. Run calibration tests against known candidates, compare the generated reports with human reviewer notes, and document how the tool is allowed to influence real hiring workflows. A practical rollout should include policy review as well as technical testing. Decide which resume fields may be processed, how long cached artifacts are retained, who can see generated reports, and how candidates can be reviewed fairly when GitHub data is missing or incomplete.