Looking for experienced software engineers to create the following software: Thousands of identical format punch cards will have been scanned and uploaded to folders on S3. Using AWS tools (EC2, etc..) and the software of your choice, analyze each card using an image analysis library to determine which holes have been punched. Create a mapping (with our help) of punched holes to fields/values, and populate the DB with data gathered from each card. Export data from the DB into an CSV file for download.
Please view the attached sample scan. There might be more complexity to the final cards. For example, some of the fields that are now write-in (e.g. "rate") might be converted to punched data. Other data might be contextual, for example, quantity (the item I punched in the sample) might populate a different field in the DB if some other part of the card is punched. In other words, there'd be a data taxonomy which would be on me to describe in some detail. The actual shape of the punch itself is not final. Also, I think we'd need to build in some error detection - perhaps some defined threshold to determine if we can conclusively read a punch, and if not, that card gets "rejected" for manual entry. Also, sample card is not scanned on final scanner, but I expect final resolution/image quality to be extremely high.