The problem: I was getting 200+ emails daily and missing important business opportunities buried in newsletters and spam. Manually sorting took hours.
The solution: A NextJS application that: - Connects to Gmail via IMAP - Uses NVIDIA's Llama-3.1-nemotron-ultra model for classification - Automatically creates and organizes emails into "AI-Priority/Needs Reply" and "AI-Priority/Business" folders - Processes emails in batches with timeout protection - Runs every 15 minutes via cron jobs
Technical details: - Built with NextJS, TypeScript, and Tailwind - NVIDIA LLM API for intelligent email analysis - Gmail IMAP integration with proper error handling - Vercel deployment with environment variable security - Comprehensive logging and monitoring
Performance: - Processes 25 emails in 33 seconds - 99% classification accuracy - Handles 200+ emails daily - Saves 2+ hours of manual sorting
The classification logic identifies: - Needs Reply: Direct questions, meeting invites, customer inquiries, personal messages - Business: Job offers, partnerships, investment opportunities, sales leads
Everything is open source with comprehensive documentation, free Vercel deployment, use of free NVidia NIM LLM API, free plan for Val Town, and customizable AI prompts.
GitHub: https://github.com/anandvc/email-prioritizer
What email automation challenges do you face? Would love feedback on the classification logic!