AI Resume Vision Screener Agent (PDF-to-Image Analysis)

Unlock Fairer & Smarter Candidate Screening with this AI Agent

This n8n workflow transforms your initial recruitment screening by acting as an AI Resume Vision Screener. It takes a candidate's resume in PDF format, converts it into an image, and then employs Google Gemini's multimodal vision model to 'read' and analyze the resume. This unique approach ensures that candidates are evaluated based on the visual content of their resume, just as a human reviewer would, making it resilient to hidden prompts or text manipulations aimed at gaming AI-driven Applicant Tracking Systems (ATS).

By focusing on visual interpretation, this AI Agent provides a more authentic assessment of a candidate's profile for a specified role (e.g., 'Plumber' in the example configuration), determining qualification for an interview based on skills and experience as presented.

Key Features & Benefits

• AI-Powered Resume Analysis: Leverages Google Gemini's advanced vision model to understand resume content from images.
• Bypass Hidden Prompts: Effectively neutralizes text-based prompts embedded in resumes to unfairly influence AI screeners.
• PDF to Image Conversion: Automatically converts resume PDFs into image format for vision model processing (uses Stirling PDF via HTTP Request in the template).
• Structured Qualification Output: Provides a clear, structured JSON output indicating if the candidate is qualified and the reason, for easy integration into subsequent hiring stages.
• Customizable Assessment: Easily adapt the analysis prompt to target specific roles and qualification criteria.
• Increased Screening Accuracy: Reduces bias and manipulation, leading to a more reliable initial candidate filter.
• Efficiency for High-Volume Recruiting: Automates a critical and time-consuming part of the hiring process, especially useful for founders and small teams.

• An n8n instance (Cloud or self-hosted).
• Google Gemini API Key with access to a vision model (e.g., gemini-1.5-pro-latest).
• Google Drive credentials (if using Google Drive to fetch resumes as per the template).
• Access to a PDF-to-Image conversion service. The template uses Stirling PDF via HTTP. For production, a private instance of Stirling PDF (or an alternative service) is highly recommended for data privacy.

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure the 'Download Resume' (Google Drive) node: Set up your Google Drive credentials and specify how the workflow should locate candidate resumes (e.g., by File ID, folder).
4. Configure the 'PDF-to-Image API' (HTTP Request) node:
   • If using Stirling PDF, ensure your instance is running and accessible. Update the URL parameter to point to your Stirling PDF API endpoint (e.g., https://your-stirling-pdf-instance.com/api/v1/convert/pdf/img).
   • Data Privacy Warning: The default template might point to a public Stirling PDF demo instance. For production use with real candidate data, you must use your own private, secure instance of Stirling PDF or an alternative trusted PDF-to-image conversion service.
5. In the 'Google Gemini Chat Model' node, select your Google Gemini credentials or create new ones using your API Key. Ensure the selected model supports vision capabilities.
6. Customize the 'Candidate Resume Analyser' (LLM Chain) node:
   • Review and modify the initial HumanMessagePromptTemplate to define the role you're hiring for and the specific assessment criteria (e.g., "Assess the given Candidate Resume for the role of Senior Software Engineer...").
7. (Optional) Adjust the 'Structured Output Parser' node if you need a different JSON schema for the output.
8. (Optional) Modify the 'Resize Converted Image' node if you need different image dimensions for the LLM, though the default (75% resize) is a good starting point for balance between detail and processing speed.
9. Activate the workflow. Test with a sample resume to ensure it processes correctly.