Maaly Nassar

Modern AI systems are increasingly used with complex, high-impact data, but traditional RAG often struggles with fragmented evidence, weak grounding, and limited explainability. This session presents an agentic AI framework for transforming disconnected genomic, biomedical, and regulatory evidence into traceable, decision-ready insight. The framework starts from patient genomic variant call files (VCFs), using bioinformatics tools to normalise and annotate variants, then links them to scientific literature, regulatory content, curated biological knowledge, ontology-grounded annotations, vector search, and knowledge-graph evidence. Rather than relying on a single generative response, specialised agents validate findings, reconcile conflicting sources, and surface explainable conclusions with provenance. Developed for variant interpretation, precision medicine, and antimicrobial resistance, the same patterns generalise to any domain where organisations must reason over heterogeneous, high-risk data.