Fetch reduces development time by 30% building in-house AI with Hugging Face on AWS SageMaker and Inferentia

AI Impact Summary

Fetch consolidated roughly 15 AI tools into an in-house ML/AI stack on AWS, leveraging Hugging Face via the AWS Marketplace Expert Acceleration Program to train and deploy transformers with Amazon SageMaker and AWS Inferentia. By replacing a black-box third-party brain, Fetch gains full visibility and control over data extraction, entity resolution, and semantic search, enabling safer, scalable rollout at high volume. A shadow pipeline validated results on full-volume traffic before promotion to production, contributing to a 30% reduction in development time and a 50% reduction in inference latency while supporting processing of 11 million receipts per day for 18 million monthly active users.