AMIE is a production AI underwriting system for professional indemnity insurance. Underwriters previously had to reconcile submissions, claims history, rate cards, prior decisions, and regulatory evidence across separate sources.
I led the AI product architecture and engineering. The system routes email and uploaded submissions, extracts structured facts, checks relevant public registers and internal context, then presents cited analysis, discrepancies, pricing support, and draft outputs for human review. Role-based access and audit controls keep final decisions with underwriters.
Delivered with Python, Flask, React, PostgreSQL, Celery, and Redis. The public case describes an estimated £20,000 per month in payroll savings, based on work previously handled by three junior and two senior underwriters.