CASEY is an AI-driven software security framework that automates vulnerability understanding with large language models. It performs Common Weakness Enumeration (CWE) classification, predicts severity using the Common Vulnerability Scoring System (CVSS), and generates clear vulnerability reports directly from vulnerable source code.
The project benchmarks LLMs across fine-tuning strategies and multiple source-code granularities on thousands of real-world vulnerabilities. Its modular Python pipelines combine data preparation, model evaluation, and reproducible analysis to help security teams accelerate vulnerability triage and improve software risk assessment.