TokenmaxxingAI token costsAgents online
GuidesLabsEventsLeaderboardToolsBriefing
Events / Boston

Microsoft Scout, Meet Tula: Extending the Same Agent Runtime to Patient Care

Boston Healthcare Cloud & AI Community session at Microsoft NERD, Cambridge. Demos the Microsoft Scout agent extended into clinical care via Tula — a self-hosted deployment pulling Epic records via SMART-on-FHIR and running local MedGemma.

Tue, Jun 30, 9:00 PMMicrosoft New England Research and Development · Cambridge · MA
Event page on MeetupMore in Boston

Why it matters

Healthcare's strict PHI constraints force local inference, auditable skill design, and open-standard integrations. Architects here see how the same agent runtime adapts when cloud API calls are restricted — a blueprint for regulated-industry deployments.

The tokenmaxxing angle

Tula runs local MedGemma inference rather than cloud APIs to avoid routing PHI to external providers — a concrete case study in routing decisions driven by compliance, not just cost. A nuanced dimension of AI FinOps often overlooked elsewhere.

From the organizers

Tula uses a self-hosted OpenClaw deployment that pulls clinical records from Epic via SMART-on-FHIR, stores them as FHIR R4 locally, and runs MedGemma for on-device inference; source code is Apache 2.0 on GitHub.