Back to work
Case Study

SupportLens AI

SupportLens AI is a support inbox copilot that helps teams review customer messages, retrieve policy context, draft replies, and keep a human in control before anything reaches the customer.

FastAPIPostgreSQLpgvectorGmail OAuthPII redaction
SupportLens AI reply workspace screenshot
Interactive Walkthrough

Step through the support workflow

SupportLens interactive screen
Inbox home with synced support messages and queue context.

Problem

Support teams jump between inboxes, policies, refund rules, shipping docs, warranty notes, and previous replies. A generic chatbot is not enough because support responses need evidence, privacy controls, and human approval.

What I Built

  • Support inbox review queue and Gmail OAuth sync
  • Knowledge upload for policies, FAQs, and product docs
  • Retrieval workflow for source-backed draft replies
  • PII redaction, encrypted token storage, and audit logs
  • Human review before Gmail draft creation

Architecture

The backend uses FastAPI, PostgreSQL, pgvector, encrypted OAuth tokens, audit logging, and protected admin routes. The frontend focuses on a review queue, message detail screen, draft workspace, source review, and privacy state.

Result

The project demonstrates practical AI engineering beyond a chatbot: retrieval, source review, privacy controls, OAuth integration, auditability, and human approval inside a useful business workflow.