Hestia

An AI-powered elder care system for Ring cameras detecting subtle distress signals with human-in-the-loop validation.

Comment

Inspiration

Elders left without 100% supervision can pass away unnoticed. Weak distress signals—such as a subtle wave, a faint call for help, or repeated blinking—are easily missed when family members or nurses step out even briefly. We built Hestia to ensure a home continues to care, even when you're away.

What it does

Hestia turns a single static room camera (placed 1.5–3m frontally) into a vigilant elder care monitor. It processes camera triggers using a custom scene engine and AI, translates subtle physical cues into distress alerts, and routes them through a human-in-the-loop validation process before triggering any alarm or SLA escalation.

How we built it

  • Hardware Integration: Connected with Ring camera feeds as the primary trigger and vision source.
  • Scene Engine: Configured scenario routing (S1–S4) based on capability profiles to evaluate subtle motion patterns.
  • Human-in-the-Loop Architecture: Implemented a human validation layer ensuring AI detections are confirmed before sounding sirens or sending high-priority alerts.
  • Backend & APIs: Built robust event processing pipelines with strict error handling and fail-safes.

Challenges we ran into

  • Distinguishing genuine, weak distress signals from ambient room movements without causing false positives.
  • Designing a low-latency pipeline to handle human validation and SLA escalations rapidly.

Accomplishments that we're proud of

  • Successfully creating a safety-first architecture where AI acts as the translator while maintaining human verification to prevent false sirens.
  • Delivering a reliable, production-ready prototype focused on real-world elder care scenarios.

What we learned

  • The critical importance of human-in-the-loop design in health and safety consumer applications.
  • Practical techniques for processing low-amplitude visual distress signals efficiently.

What's next for Hestia

  • Expanding support for additional camera models and smart home ecosystems.
  • Refining the capability profiles and scene engines for broader medical edge cases.

Built With

Share this project:

Updates

Submission history