Sun Finance automates ID extraction and fraud detection with generative AI on AWS
How a fintech slashed document processing from 20 hours to under 5 seconds.
Deep Dive
Sun Finance, a Latvian fintech processing 80,000 monthly loan applications, partnered with AWS to build an AI identity verification pipeline. Using Amazon Bedrock (Anthropic's Claude Sonnet 4), Amazon Textract, and Amazon Rekognition, they improved extraction accuracy from 79.7% to 90.8%, cut per-document costs by 91%, and reduced processing time from up to 20 hours to under 5 seconds. The solution also uses vector similarity search for fraud detection.
Key Points
- Extraction accuracy improved from 79.7% to 90.8% using a hybrid OCR + LLM approach with Amazon Bedrock (Claude Sonnet 4)
- Per-document costs reduced by 91% and processing time dropped from up to 20 hours to under 5 seconds
- Serverless fraud detection uses vector similarity search to catch 10% fraudulent requests with distinctive image patterns
Why It Matters
Demonstrates how even highly regulated fintechs can automate identity verification at scale with gen AI, unlocking lower-cost markets.