A web app for marking handwritten exams. Questions and grading criteria are registered in advance; given a PDF of answers, three AI roles, sorting, transcription and commenting, draft the feedback. Teachers edit and place the comments on screen, then export a PDF, lifting a returned test from a bare score to real feedback.
Judges valued that the project tackles teacher workload, a problem with wide impact, while keeping room for human intervention instead of handing everything to the AI. They called it grounded work whose day-to-day operation is easy to picture.
Splitting the work across three AI roles looks like the load-bearing choice. A handwritten script demands a sequence of very different judgements: sort the pages, read the writing, assess the answer. Packed into one prompt, failure becomes impossible to trace; split into roles, a human can see where it went wrong. With the teacher in control of the final comments, the result is something that can plausibly sit in a classroom, a setting with little tolerance for silent error.
See all winners of 第3回 AI Agent Hackathon with Google Cloud →
Summaries are written by this site. Project pages include demo videos, screenshots and the team's own write-up (videos may autoplay).