Hallucination, and explainability in LLM are considered to be the most important threats when deploying production grade Agentic AI model in healthcare. We present here a new architecture approach combining Graph RAG, Medical Ontologies, fined‑tuned NER, Multi Agent Models, combining both local and cloud computing for safe, secure, and efficient model deployment, using small dataset of public knowledge and real patient files in reproductive medicine in North Africa. This work is a collaborative work between public and private fertility clinics and Computer Science startup based in Tunisia. Undergoing Randomized trial is being deployed to measure the impact of Agentic AI models in managing stress and anxiety levels for patients undergoing fertility treatments.
About the speaker

CEO/Co-Founder
at Tanit Healthcare Technologies