4. Email Classifier

Overview

A service that connects to your mailbox, classifies incoming emails into customizable categories (e.g. Urgent, Finance, Personal, Spam), and surfaces a prioritized dashboard.

Primary Use Cases

• Busy professionals triaging high-volume inboxes
• Sales teams filtering lead inquiries
• Personal users separating newsletters vs. actionable mail

Key Features

• IMAP integration to fetch new messages
• Preprocessing: header + body cleaning
• Classification: multi-label BERT model
• Dashboard: React UI showing counts, filters, and drill-down
• Auto-labeling or move to folders

Tech Stack

• Frontend: React + TypeScript (dashboard, settings page)
• Backend: FastAPI (Python)
• Worker: Go service polling IMAP & pushing to classifier
• AI Models:
  • Base: distilbert-base-uncased fine-tuned on Enron / custom data
  • Optionally: roberta-base for improved accuracy

Architecture

1. Poller (Go): IMAP fetch → push raw email metadata to queue.