Design of a Web-Based Course Conversion System Using OCR and Gemini AI

In the era of digital transformation, academic services are required to be managed quickly, accurately, and in an integrated manner. However, the process of submitting course conversion requests may still be conducted manually, potentially causing recording errors, processing delays, and limited transparency of request status. This research aims to design and develop a web-based course conversion system utilizing Optical Character Recognition (OCR) and Gemini AI to automatically extract data from academic transcript documents. The development method used is Rapid Application Development (RAD), consisting of the Requirements Planning, User Design, Construction, and Cutover phases. The system was developed using React, TypeScript, Supabase, and PostgreSQL, with features including online submission, transcript upload, automatic data extraction, multi-level verification, and digital report generation. Black Box Testing showed that the main system functions operated according to the defined test scenarios. OCR testing on one academic transcript containing seven courses resulted in 20 of 21 data items being extracted correctly, with an overall accuracy of 95.24%. An error was found in the extraction of one course name, while the remaining data in the tested document were extracted correctly. These results indicate that the developed system can support the digital submission and verification of course conversion requests, while the extracted data should still be reviewed by users.