Description
Parse candidate resumes using Multimodal LLMs to detect hidden prompts and assess qualifications — fully automated with AI
📄 Automate the screening of resumes by converting PDFs into images and feeding them into vision-enabled LLMs like Gemini or GPT-4. Perfect for detecting malicious prompts or validating candidate fit
🧠 What Problem Does It Solve?
Traditional resume screening can be gamed by candidates inserting invisible prompts to manipulate AI-based ATS systems. This workflow prevents that by converting resumes into images and parsing them visually using a Multimodal AI. It offers automated and trustworthy evaluation of
📌 Candidate qualifications
📌 Hidden prompt detection
📌 Resume formatting issues
📌 Role-based suitability
⚙️ How It Works
📥 Downloads candidate CV from Google Drive
🧾 Converts PDF to image using Stirling PDF API
🖼️ Resizes the image to optimize LLM performance
🧠 Sends the image to a Multimodal LLM (e.g., Gemini 1.5 Pro or GPT-4 Vision)
📝 Returns structured JSON with qualification status and reasoning
✅ Routes qualified candidates to the next stage
✨ Key Features
📁 Works with Google Drive-synced resumes
🧠 Utilizes vision-enabled LLMs (Gemini / GPT-4)
🔍 Detects “hidden prompts” in resumes
📊 Outputs structured data like is_qualified and reason
🔄 Easily replaceable trigger (Webhook, Email, ATS)
🛠️ Built-in image resizing for speed & clarity
🧰 What You Need
✅ Google Drive access for resume intake
✅ Stirling PDF (self-hosted preferred for privacy)
✅ Multimodal LLM access (Gemini or GPT-4 Vision)
✅ OpenAI/Gemini API Key
✅ n8n instance with LangChain nodes
🚀 Setup Instructions
🔗 Connect your workflow to a Drive folder
🔧 Set up API credentials for OpenAI or Gemini and Stirling PDF
📤 Use supported formats: PDF
🖼️ Stirling PDF API → Convert to image
📐 Resize image (e.g., 75% scale)
🧠 Use LangChain Basic LLM node to analyze the image
🧾 Use Output Parser for structured results
🧪 Test on example CVs with and without hidden prompts
🔌 Integrations
Google Drive
Stirling PDF
OpenAI / Gemini
LangChain (LLM, Output Parser)
n8n
Project Link:-[https://preview--image-scan-shield.lovable.app/]
