How We Built EIR: A Secure Healthcare Agent Fleet on Google Cloud
For the past month, our team has been building EIR: Healthcare Agent Fleet , a secure multi-agent platform for patient access, recovery, scheduling, and hospital operations. We created this article specifically for the purpose of entering the All Things Agentic Hackathon . Our goal was to move beyond a single healthcare chatbot and build a real fleet of agents that can collaborate over time,…
Over the past month, our team has been developing EIR: Healthcare Agent Fleet, a secure multi-agent platform designed for patient access, recovery, scheduling, and hospital operations. This article was crafted for the All Things Agentic Hackathon, aiming to create a real fleet of agents capable of collaborating, remembering context, interacting with hospital systems, and pausing for human approval when necessary.
Hospitals often use multiple software systems, but their workflows remain fragmented. Patients may need to schedule appointments, recover post-discharge, follow up, or contact a clinician. Meanwhile, hospital teams handle reviews, inventory, scheduling, and supply operations. EIR addresses these issues by coordinating workflows through specialized agents rather than relying on a single model to handle everything.
Our agent architecture utilizes Google Agent Runtime with Google ADK and Gemini 3.5 Flash. Key components include:
- Agent Runtime Memory Bank: Allows EIR to remember safe preferences between sessions, such as a patient's preferred clinic and appointment times.
- Agent Identity and Gateway: Assigns the agent a least-privilege identity and monitors outbound calls.
- Model Armor: Adds an additional security layer.
- Cloud Logging, Trace, and Monitoring: Ensures comprehensive observability.
EIR offers separate experiences for patients, clinicians, and operations teams:
- Patients can view appointments, search availability, book/reschedule/cancel visits, and follow recovery workflows.
- Clinicians can review their schedules, inspect patients, handle recovery escalations, and resolve human review items.
- Operations teams can monitor the hospital, inspect the agent fleet, review observability, and manage supply and replenishment workflows.
Scheduling is powered by FHIR R4 on Google Cloud Healthcare API, with long-running workflows managed by Firestore, Pub/Sub, Cloud Scheduler, and Cloud Run. Recovery Episodes can span days or weeks, with EIR scheduling follow-ups, processing structured patient responses, checking adherence and risk signals, and escalating cases to clinicians when necessary.
Critical actions follow deterministic policies and require human approval. Similar processes apply to hospital supply workflows, where procurement agents draft purchase orders, but a human must approve them before execution.
We integrated Vertex AI Veo to generate short recovery education videos from pre-approved care instructions. The generative model is designed to avoid inventing diagnoses, medications, doses, or new clinical guidance.
Our key takeaway is that useful agents require more than just an LLM. Persistent memory, identity, authorization, observability, asynchronous execution, and human approval are all crucial. Building EIR taught us to treat agents as real software actors, not isolated chat interfaces.
For a live demonstration, visit https://eir-ui-658898892127.us-central1.run.app/. The project was built by Yulian Bedoya, Juan Quintero, Marycielo Berrio, and Alejandro Feria Gonzalez for the All Things Agentic Hackathon.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
