Context Intelligence Gateway for Healthcare AI: Governed Agentic Systems, MCP Tooling, and Auditable Reasoning Traces
Healthcare AI is moving from isolated chatbots toward agentic systems that retrieve clinical context, call tools, and coordinate workflows across EHR/FHIR systems, clinical notes, policies, claims, documents, and operational silos. This creates a major challenge: how can healthcare organizations use AI agents without exposing protected health information, bypassing access controls, or producing unauditable recommendations?
This talk introduces the Context Intelligence Gateway for Healthcare AI, a governance-first architecture for safe, MCP-enabled agentic systems. The gateway acts as a policy-aware control plane between users, agents, MCP servers, healthcare data sources, and reasoning models. It evaluates user identity, role, session purpose, agent identity and access, tool permissions, data sensitivity, and approval requirements before creating a safe context package for the model.
The talk also introduces auditable reasoning traces: structured records of retrieval, policy decisions, tool calls, redactions, citations, and approvals for production-ready healthcare AI.
About the speaker
Krishnendu Dasgupta
Founder & Independent Researcher at Axonvertex AI
Krishnendu Dasgupta is the Founder of AXONVERTEX AI, based in Bengaluru, India. He has 15+ years of experience across applied machine learning, healthcare AI, generative AI, graph reasoning, decentralized/private compute, cybersecurity-adjacent systems, and agentic infrastructure. His work focuses on practical, privacy-preserving, local-first AI systems with clear security boundaries, audit trails, policy controls, sandboxed tool use, and human oversight. His current work connects Free and Open Source principles with AI security and digital sovereignty: self-hosted infrastructure, open models, deterministic orchestration, privacy gates, policy registries, auditable memory, private compute, private agents, and human approval workflows for safe AI deployment. In June 2026, Krishnendu was invited by OWASP Italy as a trainer and speaker for OWASP Italy Day 2026 in Cagliari, Sardinia, where he delivered training on building, securing, and deploying AI agent swarms in a trustless decentralized ecosystem. The training focused on secure AppSec triage and remediation swarms, multi-agent systems, policy-as-code, privacy protection, PII handling, open-source foundation models, and secure deployment in decentralized environments. He has spoken and presented AI frameworks at DSC-DACH 2025, NODES 2025, Applied AI Summit 2026, Linux Foundation AI Summit Europe, CNCF Europe, Neo4j Conference, and OWASP Foundation chapter events. Since 2023, he has contributed to NIST-driven generative AI and AI risk initiatives as an independent principal investigator and researcher. He has also served as a Stanford Code in Place Section Leader in 2024, 2025, and 2026, and has contributed as a technical reviewer for more than 16 AI books published globally by Springer Nature.