SEAS SearchKG-Based Course QA
Research Project Showcase

SEAS Search

Knowledge Graph-Based Course QA System

Fine-tuning Large Language Models on GWU course data with graph-augmented reasoning for prerequisite planning and multi-hop question answering

Final project as part of CSCI_6366 Neural Networks and Deep Learning by Anurag Dhungana and Prakriti Bista

187
Bulletin Courses
CSCI & DATS
586
Schedule Instances
Spring 2026
2,828
Training Samples
Q&A pairs
Complete
Evaluation Status
26% / 34% / 38% accuracy

Project Objectives

Fine-tune LLMs for GWU course search and question answering

Build knowledge graph from course prerequisites and relationships

Enable multi-hop reasoning for complex prerequisite queries

Create reproducible research pipeline for academic course data

Technology Stack

ML/AI
Llama 3.1 8B, LoRA, Unsloth, HuggingFace
Backend
Python, Jupyter, NetworkX, spaCy
Frontend
Next.js 16, React 19, Tailwind CSS, Recharts
Visualization
react-force-graph-2d, Framer Motion