Case Studies | FinTech

Reading every receipt in seconds

Receipt OCR and VAT Extraction on Amazon SageMaker

About

A FinTech business working with Cloud Combinator on AWS. The client is anonymised at their request.

Challenge

The challenge had three focus areas.

Accuracy on the fields that matter

A VAT refund stands or falls on four values: merchant name, date, total and VAT amount. The service had to identify these correctly at the accuracy level agreed with the client, across varied real-world receipts.

Speed the user can feel

The extraction needed to be genuinely low-latency. The agreed objective was for at least ninety-five percent of individual OCR requests, measured over a rolling window, to complete in four seconds or less from the moment the API receives the request to the moment the response is returned.

A safe transition and a quality loop

The client was already doing ad-hoc processing, including Amazon Textract. The new service had to be able to run in parallel in shadow mode to compare

Solution

Cloud Combinator ran the work in two stages so the accuracy and latency targets were agreed before the production service was built.

4 seconds

Target for 95% of OCR requests (agreed objective)

6

Initial rollout markets in scope

4 fields

Merchant, date, total and VAT extracted per receipt

By the numbers:

  • 4 seconds - Target for 95% of OCR requests (agreed objective)
  • 6 - Initial rollout markets in scope
  • 4 fields - Merchant, date, total and VAT extracted per receipt
Changes

The engagement delivered a production-ready receipt extraction service on the client's own AWS account, deployed as infrastructure-as-code and accepted against the agreed accuracy and response-time measures demonstrated during user testing.

  • Real-time extraction — A SageMaker endpoint reads each receipt and returns structured tax fields through a simple HTTPS API, sized so ninety-five percent of requests complete within four seconds.
  • Built to improve — An evaluation loop retains a rotating sample, scores accuracy against labelled data, and rolls out model updates using canary or blue-green deployment to avoid disruption.
  • A clean cut-over — The service can run alongside the existing Textract path in shadow mode, so accuracy and timing are proven before the old route is retired.
  • Cost-aware by design — Short S3 retention, a HTTP API and right-sized inference keep running costs predictable as volume grows.
  • Handover — The client received the deployed service, documentation and Loom walk-throughs, with the solution running in its own AWS environment.

With the core tax fields handled reliably and quickly, the client has the foundation to expand into richer line-item itemisation and further markets, on a service that is built to keep learning from real receipts.

AWS Stack

Amazon SageMaker

For hosting the receipt OCR and key-value extraction model as a real-time endpoint.

Amazon API Gateway

For the simple, low-cost HTTPS interface into the service.

AWS Lambda

For image validation, normalisation and orchestration of the inference call.

Amazon S3

For short-lived receipt images and rotating quality samples.

Amazon CloudWatch

For latency, error-rate and throttle metrics with alarms.

AWS CloudFormation

And IAM for least-privilege, infrastructure-as-code deployment.

YOU MIGHT LIKE

Related success stories

View all case studies

Case Studies | Insights

Utilising Language Recognition, Speed, and Enhanced Security to Make Social Media a Force for Good

  • Here, we take a detailed look at how the Cloud Combinator team collaborated with another cutting-edge AI service provider that provides intelligent systems to “make social media more social” for brands and users alike.
  • Arwen AI is a UK-based startup specialising in AI solutions to manage and enhance brands’ social media interactions. Founded in 2020 by Matt McGrory, Dr. David Cole, and Joel Bailey, Arwen. AI focuses on using AI to automatically detect and remove spam, toxic comments, and other unwanted content from social media platforms.
  • The team at Arwen have three core products. ‘Moderate’ is focused on identifying and removing toxic content from social media channels. ‘Engage’ helps brands identify and engage with meaningful conversations on social media, and ‘Customize’ allows brands to apply bespoke algorithms to their channels - creating an even more effective moderation and engagement.
Read more
CONTACT US

Ready to turn AI into impact?

We'll help you spot the highest-value opportunities, reduce risk around your first AI initiative, and define a clear path to results from day one.

Why talk to us:

Outcome-driven recommendations

AWS-recognised delivery expertise

Risk-aware AI adoption

Clear next step, not a sales pitch

Start with a focused 20-minute conversation about your goals — no pressure, no commitment.

We use cookies

This website uses cookies to enhance user experience and to analyze performance and traffic on our website.

See our Privacy Policy for details.