טופס פרויקט

AutoGrade

15007303
:מספר הפרויקט
עידו שניידר, ספיר לוי, קרן סטולר
:שמות הסטודנטים המציגים
מר שמיר גד
:שם המנחה
סדנת בינה מלאכותית בחינוך: פיתוח אפליקציות
:שם הסדנה
לא מוגדר
:מסלול הסדנה
:GitHub
פוסטר
מצגת
:תקציר הפרויקט
AutoGrade addresses the critical educational need for consistent, fast, and unbiased student assessment. The application automates the intensive, time-consuming process of grading handwritten exams through a seamless digital pipeline.
Teachers upload documents of scanned tests, which are processed using MistralOCR for high-fidelity recognition of handwriting and technical layouts. AutoGrade then utilizes GPT-4o as a semantic grading agent to evaluate transcribed answers against custom, teacher-defined rubrics. By performing human-like reasoning rather than simple keyword matching, the system provides high-quality rationales and structured feedback. This allows educators to maintain rigorous standards while focusing their energy on instruction and student support.