Dev Conference by Neo4j
You only need to register once to attend all sessions.
Session Track: AI
Session Time:
Session description
This study benchmarks the performance of Tanit.AI, a fertility-specialized AI model using GraphRAG architecture, against OpenAI's GPT-3.5-Turbo and GPT-4 models in the field of artificial reproductive technology, infertility, and human reproduction. We utilized 400 publicly available guidelines and publications, transforming them into a knowledge graph enriched with medical ontologies. We then applied embeddings and vector indexes for natural language querying via a large language model. The model's accuracy was tested using 180 YES/NO questions from a reproductive medicine master's program. Tanit.AI demonstrated competitive accuracy (75.42%) and precision (82.61%) compared to GPT-4 and GPT-3.5-Turbo. Additionally, Tanit.AI provided explainable answers with references to trusted sources. GraphRAG offers promising results for specialized LLMs in the medical field, enhancing knowledge access and decision-making for doctors and patients. This study paves the way for precision medicine and graph AI in healthcare.
Founder & CEO, Tanit Healthcare Technologies
Kais Zhioua is a Board Member at Tunisia Health Alliance, Clinique la Rose and Fertillia and founder of Tanit Healthcare Technologies, an personal companion towards parenthood. He works as Chief Strategy and innovation at Clinique la Rose and as the Executive Coordinator of Africa Healthcare Investment Summit. He started working in the healthcare sector since 2013. In 2016 he started working in Tunisia, with an extensive focus on the Fertility treatments in Africa. His experience encompassed, among others, strategic development, digital marketing, market research, international cooperation, advocacy, fundraising, healthcare infrastructure development and recently AI. Kais was born in Tunis, Tunisia in 1988. He graduated in 2015 from Krannert Management School, Purdue University and EM Lyon in Global Entrepreneurship and Innovation.