hiring-agent: Score Resumes with AI and GitHub Signals

Overview

Hiring Agent is a Python pipeline that extracts information from PDF resumes, optionally adds public GitHub profile and repository signals, and evaluates candidates against a role-specific rubric. Its intended purpose is to help prioritize which resumes receive attention, rather than replace an applicant tracking system or make hiring decisions on its own.

It is most relevant to developers and teams who want to inspect or adapt an LLM-based resume evaluation workflow. The project supports local models through Ollama as well as Google Gemini, and exposes its evaluation prompts and role configuration for scrutiny and customization.

Key Features

• Converts PDF resume content into structured data using LLM-assisted, section-based extraction.
• Enriches profiles with GitHub account and repository information when a profile is found in the resume.
• Scores candidates against configurable role-specific categories, weights, bonuses, and deductions.
• Produces category scores and supporting evidence through the evaluation pipeline.
• Supports Ollama-hosted local models and Google Gemini.
• Provides role scaffolding so users can create and edit different evaluation rubrics.
• Offers optional caching and CSV export in development mode.

Use Cases

• A recruiting or engineering team wants to prioritize a large resume queue for human review using a configurable rubric.
• A developer wants to prototype a resume parsing and scoring pipeline with local or hosted LLMs.
• A hiring team wants to inspect how rubric categories and prompts affect evaluations before adapting them to a role.
• A candidate or researcher wants to examine how resume details and public GitHub signals are represented in an automated evaluation.

Project Facts

• Language: Python
• License: MIT
• Stars: 7.3k
• Forks: 1.4k
• Archived: No

Getting Started

Requires Python 3.11+ and either Ollama or a Google Gemini API key. A minimal setup is:

git clone https://github.com/interviewstreet/hiring-agent
cd hiring-agent
python -m venv .venv
pip install -r requirements.txt

Configure a model and provider in the environment, then run python score.py ./resume/sample.pdf --role software_engineering_intern. See the README for setup details and role configuration.

Considerations

• Scores depend on LLM output, so repeated evaluations may vary; the README links to outside analyses discussing consistency concerns.
• GitHub enrichment can favor candidates with public repositories and may not reflect work kept in private repositories.
• The default local model is a demo configuration, not the model described for production intern resume evaluation in the README.
• Resume filtering has consequential fairness and privacy implications. Treat results as prioritization signals and retain human review rather than using scores as a sole decision basis.
• Running requires Python 3.11+ and an LLM backend; Gemini use requires an API key, while local inference depends on the selected Ollama model and available hardware.