AI-powered malware detection system using a hybrid CNN + BERT model, achieving ~88% accuracy on real-world datasets. The project demonstrates end-to-end skills in feature extraction, deep learning, and cybersecurity, effectively detecting metamorphic malware that evades traditional antivirus tools.

A production-ready Retrieval-Augmented Generation (RAG) assistant designed for customer support, leveraging hybrid search, LLM-based generation, and policy enforcement for accurate, citation-based responses. Includes features for real-time data handling, PII masking, and SLA compliance in energy and telecom domains.