Summary

Uses LLMs for enterprise security automation including vulnerability detection, code scanning, correct code generation, and fine-tuning. Achieves improved F1-score accuracy from 0.21 to 0.60 for vulnerability detection. Demonstrates network configuration facilitation, GitHub issue analysis, and code-specific sub-model derivation using efficient pruning.