dtax
An educational prototype that turns a sample patient's spoken account into a symptom summary and a slip for clinician review.
Elastic Hackathon · 28 February 2026
Patient Intake
We built dtax to explore the step between a patient describing symptoms and a clinician reviewing the visit. In the demonstration, users select a sample patient, take part in a voice interview, follow the transcript, and review a generated symptom summary and triage slip.
The slip can be printed or shared, while a clinician dashboard shows the handoff details and moves visits through waiting, consultation, and completion. This is an educational prototype with sample patients and generated urgency labels, not a validated clinical triage service.
Turning an Interview Into a Clinical Handoff
Separating Conversation From Ticket Generation
The voice interview and the final slip use different model tasks. Nova 2 Sonic supports the spoken exchange through Bedrock; Nova 2 Lite takes conversation history and patient context to generate a structured triage ticket. This keeps the conversational account available as input while producing a separate summary for the handoff, as described in the application architecture.
Conversation and Handoff
| Component | Responsibility |
|---|---|
| Nova 2 Sonic | Supports the voice interview through Amazon Bedrock. |
| Nova 2 Lite | Turns conversation history and patient context into a triage ticket. |
| Amplify and DynamoDB | Organize patient, conversation, ticket, and queue records. |
| Clinician dashboard | Shows handoff details and moves visits from waiting to consultation to completed. |
Moving the Visit Through the Queue
Patient, conversation, ticket, and queue records represent different parts of the visit. The clinician opens the handoff details and changes the queue status from waiting to consultation to completed. Generating the slip supplies the information to review; updating the queue records where the visit is in the workflow.
Prototype and Talk
My presentation at the Amazon Web Services (AWS) User Group covered the team's build and deployment process, including the choices needed to fit voice interaction, data modeling, and a working demonstration into the hackathon.