Swift Grader
LLM-powered essay grading system built in 3 days at HackFax x PatriotHacks 2026.
Project Snapshot
Built an AI-powered essay grading backend that evaluates student submissions against teacher-defined rubrics using a multimodal LLM. Engineered multi-format file processing and a structured grading pipeline, reducing manual grading time by approximately 80%.
Problem
Manual essay grading is time-consuming and inconsistent. Teachers need a tool that can evaluate submissions against custom rubrics while preserving human oversight for final review.
Approach
- Integrated a multimodal LLM model for rubric-based essay evaluation
- Engineered multi-format file processing pipeline supporting PDF, image, and text ingestion
- Designed FastAPI backend generating structured, criterion-level feedback
- Maintained human-in-the-loop review for quality assurance
Tech Stack
Python, FastAPI, LLM API, PDF/Image processing libraries
Results
- Reduced manual grading time by ~80%
- Delivered a production-ready prototype in 3 days under hackathon constraints
- Structured criterion-level feedback for each submission