APPLeSEEd Lab
Software Engineering Research Intern
June 2025 - Sept. 2025University of Central Florida, Orlando, FL
Fine-tuned CodeBERT on Linux kernel commit data to classify security risk in software engineering research.
PythonCodeBERTPyTorchMachine LearningCVE DataGit
Highlights
- Fine-tuned CodeBERT on 10,000+ Linux kernel commits to build a commit risk classifier.
- Achieved F1 = 0.69 on unseen test data, demonstrating feasibility of AI-assisted commit analysis for software engineering.
- Labeled the dataset by linking git commits to public CVE records, enabling supervised training for security-risk classification.