Automated Cyber Threat Intelligence Extraction réalisé par un freelance en Tunisie | Tunisie Freelance
Automated Cyber Threat Intelligence Extraction
web-development
Rôle: Machine Learning Engineer Intern
Developed an end-to-end LLM pipeline to extract and structure Cyber Threat Intelligence (CTI) from confidential PDF reports into STIX 2.1.
Built and annotated a 4,738-example dataset and fine-tuned Mistral-7B and LLaMA 3.1 8B using QLoRA, 4-bit quantization, and PyTorch, achieving 0.7253 Macro F1.
Integrated MITRE ATT&CK/MISP knowledge graphs, relation extraction, confidence scoring, and XAI for traceable CTI generation.
Built an Angular + FastAPI application and deployed the solution on AWS (EC2, ECR, S3, CloudFront, CloudWatch) using Docker.