Workforce & Caregiver Logistics Engine · Target event: Healthcare Hack NYC
A single conversational voice AI that answers the phone for a home-health agency and handles the four jobs that eat scheduler time: back-filling call-outs, intaking & matching new patients, logging medication compliance, and bridging language + crisis escalations — all while holding perfect state across interruptions.
This file is the entry point. The full product and engineering documentation lives in
docs/.
| Doc | What's inside |
|---|---|
| docs/README.md | Navigation hub + suggested reading paths |
| 00 · Executive Summary | Problem, solution, value proposition, business metrics |
| 01 · Product Requirements | Personas, user stories, functional requirements, scope, success metrics |
| 02 · Architecture | Orchestrator + SubWorkflows, Dual-Agent core, barge-in, diagrams |
| 03 · SubWorkflows | The four specialized flows (one spec each) |
| 04 · Data Model | Neo4j graph schema, calendars, audit logs, checklists |
| 05 · Integrations | Twilio Voice/SMS/Video, ElevenLabs, Neo4j |
| 06 · Non-Functional Requirements | Latency, HIPAA, reliability, security, i18n, observability |
| 07 · Roadmap & Execution Plan | Hackathon build order, milestones, demo script |
| 08 · Glossary | Terms, personas, acronyms |
Home-health agencies spend ~27% of their operating budget on non-clinical admin labor — mostly manual phone coordination. When a caregiver calls out at 6:00 AM, a human scheduler burns 45 minutes dialing backups while billable shifts go unstaffed. The Healthcare Omni-Agent replaces that scramble: a Main Orchestrator Agent answers every Twilio call, recognizes the caller, detects intent, and routes into one of four SubWorkflows. A Dual-Agent core (an Active voice agent + a Passive context ledger) keeps conversation state flawless even when callers interrupt or switch topics mid-call.
Draft — hackathon build. See 07 · Roadmap & Execution Plan
for current scope and build order.