From uploaded documents to ready-made course content

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had three focus areas that the build needed to address.

An user uploads a document to Amazon S3 through an Amazon API Gateway and AWS Lambda using a pre-signed URL, with every stage logged to Amazon DynamoDB. The upload triggers an AWS Step Functions workflow that runs an asynchronous Amazon Textract job to extract the text, using an Amazon SQS queue to throttle requests within Textract's service limits and Amazon SNS to signal completion. Extracted text is stored back in Amazon S3.

Cloud Combinator delivered a working, secure pipeline that converts uploaded documents into structured Q&A course content automatically, meeting the client's quality bar and deployed into their own AWS account as infrastructure as code.

• Automated content — Uploaded documents are turned into question-and-answer benchmarks in JSON, removing the manual authoring step.
• Reliable extraction — Amazon Textract reads varied document types, and the project tested direct model ingestion to widen the file types the product can accept.
• Built to scale — Amazon SQS throttling and Step Functions orchestration keep processing within service limits so high upload volumes are handled without lost work.
• Fully traceable — Every stage of each document's journey is logged in Amazon DynamoDB, giving clear visibility of processing status.
• Handover — Cloud Combinator delivered the solution into the client's AWS account with a demonstration, materials and documentation.

With the pipeline proven, the natural next step (projected) is to broaden the supported document types, integrate the output directly into the client's product, and run a Well-Architected review to take the solution to full production standard.