Inspiration Low-code/No-code services like Microsoft power platform. When a Power platform can enable app development, automation of workflows, thought about how a OCR based application for Document processing can be built with similar Low-code/No-code
What it does A Low-Code/No-Code based Product which is scalable (hosted on Azure Kubernetes with CI CD) to enable Intelligent Data extraction to accelerate data entry for Loan Officers/other Finance product users using Computer Vision of Azure, Self Service Bot with QnA Maker for answering Financial related Questions, Prediction for Mortgage approval/House pricing - Azure ML model
How we built it Leveraging the power of Vision API of Azure platform, processed the document and store the co-ordinates of one baseline image. Similar structured documents can be uploaded which inturn get stored in Microsoft Blob storage. Using the co-ordinates saved in Cosmos DB, we process multiple documents in batches and extract the data. Also, we have a Self service bot built using QnA Maker. Solution is hosted on Azure Kubernetes Services. Data is stored in Cosmos DB with API for MongoDB.
Challenges we ran into High Cost of Cosmos Database with limited resources created pressure for DB storage
Accomplishments that we're proud of Successfully implemented the Low code / No code way of OCR product which eliminates coding effort and any end user can leverage this product for defining his model himself and start uploading multiple documents in batches
What we learned Power of Vision API, Cosmos DB usage
What's next for Intelligent Extractor Integrate more cognitive services like Face API, Speech to text convertor for further taking this product to next level
Log in or sign up for Devpost to join the conversation.