Home IQ

HomeIQ turns Fire TV into an AI energy command center with Alexa voice control and Family Guardian elder-care protection that detects kitchen hazards and cuts power bills from your living room screen.

Comment

💡 Inspiration

Every year, global households waste over 30% of their electricity on standby phantom loads and unoptimized peak-hour tariffs. But an even deeper, more emotional problem struck closer to home: the silent vulnerability of our elders living independently.

A few months ago, an elderly family member in Bengaluru left their kitchen gas stove ignited late at night after brewing tea. Hours passed with zero kitchen movement, posing a life-threatening fire hazard that was only caught by sheer luck.

We asked ourselves: Why are smart home dashboards just cold grids of toggles and kilowatt graphs? Why aren’t the biggest, brightest screens in our living rooms—our Amazon Fire TVs—protecting the people we love while actively saving our households money?

This led to HomeIQ: a 10-foot ambient smart home command center for Amazon Fire TV and Alexa, transforming raw telemetry into proactive AI coaching with Amazon Bedrock (Claude 3 Haiku) and introducing Family Guardian—an emotional safety net that safeguards our families.

⚡ What It Does

HomeIQ bridges Amazon Echo and Fire TV into a unified smart home intelligence ecosystem:

  1. Ambient 10-Foot Command Center: Displays real-time power telemetry across 13 devices in 4 rooms, interactive climate controls, dynamic camera feeds, and circuit tripwires designed natively for the D-pad remote.
  2. Dynamic Bézier Spline Analytics: Visualizes household power draw across Today, Week, Month, and Year intervals using cubic Bézier smoothing to isolate peak-tariff demand spikes.
  3. "IQ" AI Home Companion (Amazon Bedrock): Powered by Anthropic's Claude 3 Haiku via Bedrock, "IQ" acts as an empathetic personal energy coach. It generates time-of-day contextual advice, answers complex billing questions, and provides interactive two-way controls (e.g., Turn OFF All Lights).
  4. Family Guardian (Elder Care & Child Safety):
    • Elder Care Mode: Detects abnormal late-night routines, such as an active kitchen stove at 2:00 AM with zero movement for 180 minutes.
    • Slide-In Hazard Alert: Slides down an urgent notification banner directly on the Fire TV with an animated [🛑 Auto-Shutoff in 60s] emergency circuit breaker.
    • Child Safety & Lone Occupant: Alerts if exterior smart locks open during school hours or if a solitary resident shows extended inactivity.
  5. Cross-Device Alexa Voice Trigger: Hands-free Echo queries seamlessly deep-link directly into the Fire TV app ("Alexa, show my energy report" ➔ homeiq://energy_report).
  6. Energy Gamification & Streaks: Daily challenges (Peak Tariff Defender, Sleep Saver), weekly family leaderboards, and an active 🔥 7-Day Eco Streak that turns conservation into a fun daily household routine.
  7. Resident Authentication & Fast Demo Profiles: Complete Sign Up, Login, and 1-click remote demo profiles (Venkat Reddy Admin, Alex Johnson Resident, Priya Caretaker).

📐 The Energy Mathematics

HomeIQ calculates real-time household power consumption and dynamically projects time-of-day tariff burdens using continuous integration:

$$\text{Total Cost} = \int_{0}^{T} P(t) \cdot C_{\text{tariff}}(t) \, dt$$

Where:

  • \( P(t) = \sum_{i=1}^{n} V(t) \cdot I_i(t) \cdot \cos(\theta_i) \) is instantaneous aggregate real power across active circuits.
  • \( C_{\text{tariff}}(t) \) models time-of-day utility rates (surging during 5:00 PM – 9:00 PM peak hours).

To render 60 FPS charts on Fire TV hardware without GPU frame drops, points are mapped along a cubic Bézier spline curve:

$$B(t) = (1-t)^3 P_0 + 3(1-t)^2 t P_1 + 3(1-t) t^2 P_2 + t^3 P_3, \quad t \in [0, 1]$$

🛠️ How We Built It

  • Fire TV & Android TV App: Built natively in Kotlin 1.9.22 using Jetpack Compose for TV, AndroidX Leanback, and Material 3 with unidirectional StateFlow (MVVM).
  • Generative AI Core: Integrated Amazon Bedrock (Claude 3 Haiku) with an empathetic system prompt persona, providing time-aware proactive insights and actionable quick-action intent generation.
  • Alexa Custom Skill & AWS Lambda: Created an ASK custom skill (alexa-skill/skill.json and interactionModel.json) with an AWS Lambda backend that constructs and dispatches Android deep link URIs (homeiq://) to the Fire TV Leanback task stack.
  • Safety Anomaly Engine: Implemented AnomalyDetector.kt, a rule-based inference engine that monitors live appliance wattage against ambient PIR motion logs and time-of-day boundaries.
  • 10-Foot Design System: Customized neon cyan illumination #38BDF8, deep midnight navy glass #070B16, and scale animations (1.05x) that strictly decouple D-pad focus states from toggles.

🧗 Challenges We Ran Into

  1. Fire TV 1080p Dialog Width Constraints: Jetpack Compose Dialog components initially defaulted to mobile portrait widths (560dp max). We resolved this on 1080p landscape TV displays using DialogProperties(usePlatformDefaultWidth = false) to achieve full-screen modal overlays.
  2. D-Pad Focus State Coupling: Early builds experienced "pre-selected" button highlights because Compose cursor focus and persistent toggle state shared visual indicators. We completely decoupled focus cursors from selection states.
  3. Cross-Device Deep Link Task Affinity: Triggering Alexa intents while the app was running in the background initially caused activity restarts. We solved this by configuring launchMode="singleTask" in AndroidManifest.xml and routing payloads through onNewIntent.
  4. We documented our complete technical journey across 7 detailed points in our official *Developer Friction Log*.

🏆 Accomplishments That We're Proud Of

  • Bridging Two Ecosystems: Connecting an Amazon Echo voice prompt to a live animated visual state on Fire TV in less than 400 milliseconds.
  • The Emotional Hook: Moving beyond plain telemetry into a genuine life-safety platform that can prevent house fires and protect vulnerable elders living alone.
  • Living Room AI Persona: "IQ" doesn't just return markdown; it speaks with warmth, addresses family members by name, and provides one-click action buttons right on the TV screen.
  • Buttery 60 FPS Charts: Rendering Catmull-Rom spline curves with neon glow gradients smoothly on entry-level Fire TV Stick hardware.

📚 What We Learned

  • How to architect 10-foot user interfaces that feel intuitive when operated purely via a directional pad and select button from 10 feet away.
  • How to structure Amazon Bedrock prompt schemas to return structured, deterministic action payloads alongside natural language responses.
  • The power of developer friction logs in discovering edge cases within Android TV Compose and Leanback libraries.

🔮 What's Next for HomeIQ

  • Matter & Thread Protocol Integration: Direct peer-to-peer pairing with smart circuit breakers and zigbee sub-meters.
  • Camera Computer Vision (Amazon Rekognition): Localized edge fall detection on security camera streams to alert family members immediately if an elder falls.
  • Community Grid Micro-Trading: Allowing neighborhoods with solar panels and batteries to exchange excess green energy peer-to-peer during peak grid alerts.

Built With

  • ai-coach
  • alexa
  • alexa-skills-kit
  • amazon-bedrock
  • ambient-computing
  • android-tv
  • aws-lambda
  • claude-3-haiku
  • deep-linking
  • elder-care
  • energy-analytics
  • fire-tv
  • generative-ai
  • iot
  • jetpack-compose
  • json
  • kotlin
  • leanback
  • material-design-3
  • mvvm
  • node.js
  • restapi
  • smart-home
  • stateflow
Share this project:

Updates

Submission history