CareSentinel+

Autonomous ambient care & home safety OS for solo seniors, fusing Alexa+ MCP, Ring Vision, and AWS Cloud into a proactive life-safety guardian.

Comment

Inspiration

Over 58 million elderly individuals live alone globally, a demographic projected to surpass 80 million by 2030. Today's smart home devices are entirely passive. They assume an able-bodied resident who can walk effortlessly, see clearly, grab a smartphone, and navigate complex touchscreen apps.

When an 78-year-old falls in a hallway, slips in a bathroom, or suffers sudden hypoglycemic dizziness, passive technology fails completely:

  • Seniors cannot reach a phone resting on a counter.
  • If they shout for help, traditional voice assistants merely recite web search results or declare they did not understand.
  • Worst of all, even if an alarm eventually sounds, the front deadbolt remains locked from the inside. Emergency first responders lose the critical Golden Hour of trauma intervention attempting to breach reinforced doors.
  • Inadvertent medication double-dosing costs healthcare systems over 34 billion dollars annually in preventable hospitalizations, driven by mild cognitive impairment and unverified manual tracking.

We built CareSentinel+ to transform smart home devices from passive gadgets into an autonomous, proactive ambient guardian that protects elderly independence, eliminates dangerous medication confusion, and provides distant families with complete peace of mind.

What it does

CareSentinel+ is an autonomous ambient care and home safety operating system that fuses Alexa+ (Voice Brain via Model Context Protocol), Ring (Smart Vision and Access Simulator), and AWS Cloud (Bedrock and SNS) into a proactive life-safety guardian:

  1. Autonomous Emergency Fall Triage and Auto-Unlock: When an elder falls or experiences acute distress (for example: Alexa, I fell and hurt my hip), CareSentinel+ executes clinical triage, automatically disengages the smart deadbolt so paramedics can enter immediately without forcing entry, and dispatches high-priority SMS alerts with exact home telemetry to family via AWS SNS.

  2. Geriatric Double-Dose Interception: Maintains an immutable medication adherence timeline. If a senior forgets and tries taking medication twice, Alexa actively intervenes: Hold on! You already took Metformin today at 8:15 AM. Please do not take an extra dose.

  3. Proactive Perimeter Defense: Analyzes Ring camera motion at night (such as 11:32 PM). When couriers arrive, the outdoor Ring speaker instructs: Please leave package on doorstep, resident is resting. The deadbolt remains engaged, keeping the senior safe inside from doorstep danger.

  4. Echo Show 15 Smart Display Console: A high-contrast, multi-modal console featuring continuous hands-free voice listening (Alexa wake word), real-time medication checklist synchronization, and a live MCP JSON-RPC telemetry inspector.

How we built it

CareSentinel+ was engineered from scratch adhering strictly to Amazon's latest hackathon specifications:

  • Backend / MCP Server (Python and FastAPI): Implemented Amazon's official Streamable HTTP MCP Specification (2025-11-25+) with Server-Sent Events (/mcp/sse) and JSON-RPC 2.0 message handling (/mcp/messages). Engineered 5 strictly-typed Pydantic tools: analyze_visitor_intent, control_smart_lock, emergency_health_triage, caregiver_dispatcher, and medication_schedule_logger.

  • AI Reasoning and Cloud Layer (AWS Bedrock and AWS SNS): Integrated Claude 3.5 Sonnet on AWS Bedrock via boto3 with high-fidelity deterministic local heuristic fallbacks, and AWS SNS for emergency multi-channel alert dispatching.

  • Frontend / Echo Show 15 Simulator (Next.js 14, TypeScript and Tailwind CSS): Designed an Apple/Linear-grade dark console simulating an Echo Show 15 smart display, powered by browser-native Web Speech API for continuous hands-free voice recognition and speech synthesis.

  • Automated Verification: Engineered a standalone test suite (test_suite.py) verifying all 5 MCP tools and end-to-end orchestration with 6 out of 6 tests passing (100%).

  • Deployment: Zero-friction cloud deployment on Render (Backend MCP Server) and Vercel (Next.js Frontend).

Challenges we ran into

  • Streamable HTTP Heartbeat Timing: Implementing the latest 2025-11-25+ Streamable HTTP MCP specification over Server-Sent Events (SSE) without connection timeouts through proxy layers required engineering a dedicated 15-second background keep-alive ping loop.

  • Zero-Hallucination Safety Air-Gap: Ensuring that LLMs never possess direct destructive actuation privileges (like unlocking doors or prescribing drugs). We resolved this by building a strict deterministic MCP schema air-gap where all LLM calls must pass hardcoded clinical and physical invariants.

  • Multi-Modal Audio Feedback Loops: Managing browser-native continuous speech recognition so that Alexa speaking via SpeechSynthesis does not trigger its own wake word or microphone input.

Accomplishments that we're proud of

  • True Cross-Device Autonomy: Unifying Ring vision, smart deadbolts, and Alexa+ voice into synchronous actions without requiring manual routines or user clicks.
  • 100% Deterministic Safety: Zero hallucinations in medical dosages and emergency door actuation.
  • Automated Test Suite: 6 out of 6 comprehensive automated tests passing with 100% code integrity.
  • Zero-Hardware Accessibility: Fully testable by any evaluator globally via our 1-click live deployment without requiring 500 dollars or more in physical hardware.
  • Developer Ecosystem Feedback: Authored an in-depth, constructive developer feedback report (FRICTION_LOG.md) analyzing Amazon's developer tooling and Streamable HTTP protocol.

What we learned

  • How to build self-hosted Model Context Protocol (MCP) servers conforming to Amazon's latest Streamable HTTP standard.
  • The critical architectural difference between reactive chatbots and proactive ambient intelligence.
  • Best practices for multi-modal smart display UX combining glanceable visual telemetry with gentle voice summaries.

What's next for CareSentinel+

  • Acoustic Fall Sound Detection: Adding ambient acoustic classification to detect loud thuds or glass breaking even if the resident is unable to speak.
  • Amazon Pharmacy Auto-Refill: Integrating Amazon Pharmacy APIs so when adherence logging detects low pill counts, refill orders are queued automatically.
  • Wearable Sensor Integration: Syncing with Apple Watch and Galaxy Watch for real-time heart rate spikes and ECG telemetry.
  • B2B2C Insurance Partnerships: Partnering with Medicare and health insurance providers who incentivize ambient safety systems to reduce 34 billion dollars in preventable fall and overdose hospitalizations.

Built With

Share this project:

Updates

Submission history