AI/ML Engineer — Gemini Vision Image Selection Pipeline

We are building an AI-powered faceless YouTube video creation platform at makeavideos.com. We need a fast, experienced AI/ML engineer to build an automated image selection system using Gemini Vision. The problem: Our platform auto-generates documentary videos. We use Scrapedog to fetch stock images per script section but results are often contextually wrong — requiring manual user review. The goal: Build a Gemini Vision scoring pipeline that: Receives script section text + 10-20 candidate images Scores each image for contextual relevance Auto-selects the best match per section Flags low-confidence sections only Eliminates the manual review step entirely Tech stack: Gemini Flash Vision API Node.js / Railway backend Existing pipeline: Claude + Scrapedog + Cartesia TTS + Remotion + AWS Lambda Requirements: Strong Gemini Vision or GPT-4V experience Node.js backend Production AI pipeline experience — no beginners Available to start immediately This is a focused, well-scoped task — not a complex project. A senior AI engineer should complete this in 3-5 days. Budget: $300-600 Timeline: Maximum 1 week, prefer 3-5 days Start: Immediate