AI/ML to Enhance and Streamline Manual/TA Grading

This project develops an intelligent clustering system that groups student submissions by similar patterns, algorithmic approaches, and correctness levels. The system enables TAs to efficiently manage large-scale grading while maintaining fairness and consistency. Problem: TAs face repetitive, time-consuming grading of similar submissions with risk of inconsistency and human error, especially in large courses (100+ students). Solution: Automated clustering of submissions using hierarchical agglomerative clustering with interactive dendrogram visualization. TAs can merge/split/reject clusters, bulk apply feedback templates, and flexibly grade (bulk or individual). Key Deliverables: - Feature extraction engine (68-dim for code, 389-dim for text) - Hierarchical clustering with Ward linkage - Interactive dendrogram UI for TA manipulation - Bulk feedback application system with per-student customization - Flexible grading modes (bulk, individual, hybrid) - Comprehensive testing and documentation Expected Impact: 80-90% reduction in grading time while improving consistency and feedback quality. Time savings: 45 submissions graded in 15 minutes vs 225 minutes without clustering.