John Snow Labs 56d ago

Clinical AI fails where context matters most: across encounters, modalities, and time. Vector search retrieves fragments. Clinical reasoning requires relationships. Join Krishnendu Dasgupta (Founder, AXONVERTEK AI; CTO & Co-Founder, TrialBridge AI) for: G.A.M.E.R.S: Graph Agents with Multimodal Entities and Reasoning Schemas in Clinical World Clinical data is not a modelling problem. It is a structure problem. Radiology, labs, prescriptions, and notes are processed in isolation, breaking the patient narrative and introducing hallucinated mappings, missing temporal signals, and incorrect reasoning paths. G.A.M.E.R.S proposes a graph-native approach where entities, not tokens, become the unit of reasoning, enabling consistent reasoning across multimodal clinical data. Timestamps 00:00 Why clinical AI loses context across modalities 02:20 Limits of RAG and vector-only retrieval 05:10 G.A.M.E.R.S architecture: graph + entities + reasoning 08:40 Multimodal ingestion and graph-based memory 12:30 Agent workflows: drug safety and cohort alerts 16:20 Explainability and graph-based reasoning paths 20:10 Demo: graph traversal and clinical decision flow – Graph encodes relationships, not just similarity – Reasoning happens across entities, not isolated documents – Multimodal inputs are fused into persistent clinical memory – Evidence paths are traceable through multi-hop traversal Instead of flat retrieval, the system operates across ICD hierarchies, RxNorm mappings, lab trends, and encounter timelines, enabling decisions grounded in longitudinal context. Agent workflows extend this into practice: – Drug safety checks using ingredient-level reasoning and lab signals – Cohort alerts based on evolving patient state – Human approval enforced at decision boundaries This is not about better generation. It is about restoring structure to clinical reasoning. 📌 Applied Healthcare AI Summit 2026 — what actually works in real-world healthcare AI, from pilots to production systems. #ClinicalAI #GraphAI #AIAgents #MultimodalAI #HealthcareAI #RAG #AIArchitecture