We present SAVE-ME, a novel tool for automated vulnerability detection in Erlang codebases that integrates traditional static code analysis with advanced machine learning techniques. By fine-tuning CodeBERT, a transformer-based language model, on Erlang functionlevel data labeled using the InfERL tool, SAVE-ME achieves functionlevel vulnerability classification tailored to the unique characteristics of functional programming and concurrency in Erlang. The approach enhances the precision of static analysis by using deep learning to reduce false positives and uncover subtle security issues overlooked by conventional tools. Developed as part of the RESCALE project, SAVE-ME contributes to improving the security of software supply chains through AIassisted code auditing. This work demonstrates the potential of combining language models and static analysis for building more secure Erlangbased systems and outlines directions for future enhancements, including multi-label classification and deeper integration into CI/CD pipelines.