CareGraph AI
An autonomous voice medical agent built on a graph database for emergency triage. It performs real-time symptom matching, allergy checks, hospital bed reservation and physician notification.
Timeline (last 90 days)
Developer Roshni Mishra built CareGraph AI in FalkorDB Graph Hacks, an autonomous voice medical triage agent based on the FalkorDB graph database.
CareGraph AI can execute 5-hop relationship queries on FalkorDB in about 14 milliseconds, completing the entire process from symptom matching to locating ICU beds and specialists.
The medical knowledge graph nodes of CareGraph AI include patients, symptoms, diseases, medications, hospitals, doctors, and beds, and the edges represent relationships such as "exhibits", "indicates", "contraindicates", "employs", and "has beds".
The agent's decision history is stored in the graph as AgentSession nodes, forming queryable structured memory rather than a messy chat log.
After the patient describes symptoms by voice, the audio is converted to text by Deepgram, and the agent executes a 5-hop Cypher query that completes diagnosis, contraindication checking, hospital selection, doctor assignment, and bed location in a single pass.
CareGraph AI uses Murf Falcon speech synthesis (hi-IN-shweta voice), with speech synthesis latency below 250 milliseconds, supporting Hindi and English.
CareGraph AI initiates autonomous PSTN outbound calls to hospital emergency desks via the Twilio REST API, announcing patient information, allergy warnings, and on-duty doctor notifications.
CareGraph AI creates an AgentSession node after the query returns, and automatically triggers a follow-up call 24 hours later to check on the patient's recovery.
All Facts (8)
Developer Roshni Mishra built CareGraph AI in FalkorDB Graph Hacks, an autonomous voice medical triage agent based on the FalkorDB graph database.
50%UnverifiedCareGraph AI can execute 5-hop relationship queries on FalkorDB in about 14 milliseconds, completing the entire process from symptom matching to locating ICU beds and specialists.
50%UnverifiedThe medical knowledge graph nodes of CareGraph AI include patients, symptoms, diseases, medications, hospitals, doctors, and beds, and the edges represent relationships such as "exhibits", "indicates", "contraindicates", "employs", and "has beds".
50%UnverifiedThe agent's decision history is stored in the graph as AgentSession nodes, forming queryable structured memory rather than a messy chat log.
50%UnverifiedAfter the patient describes symptoms by voice, the audio is converted to text by Deepgram, and the agent executes a 5-hop Cypher query that completes diagnosis, contraindication checking, hospital selection, doctor assignment, and bed location in a single pass.
50%UnverifiedCareGraph AI uses Murf Falcon speech synthesis (hi-IN-shweta voice), with speech synthesis latency below 250 milliseconds, supporting Hindi and English.
50%UnverifiedCareGraph AI initiates autonomous PSTN outbound calls to hospital emergency desks via the Twilio REST API, announcing patient information, allergy warnings, and on-duty doctor notifications.
50%UnverifiedCareGraph AI creates an AgentSession node after the query returns, and automatically triggers a follow-up call 24 hours later to check on the patient's recovery.
50%