Fetch consolidates AI tools with Hugging Face on AWS, cuts development time by 30%

AI Impact Summary

Fetch replaced a third-party AI brain with in-house ML pipelines on AWS, leveraging Hugging Face transformers to process receipts data and analytics at scale for 18 million monthly active users and 11 million receipts per day. By engaging Hugging Face through the AWS Marketplace Expert Acceleration Program and running models on Amazon SageMaker with AWS Inferentia, Fetch cut development time by 30% and halved inference latency while validating results via a shadow pipeline to avoid disrupting live processing. The move increases data granularity and control over insights for business partners and reduces reliance on external AI vendors, lowering operational risk.