Inspiration
Over 1 billion people worldwide live in regions characterized by abundant solar radiation but burdened by volatile power grids and escalating utility costs. Despite the clear economic and ecological benefits of distributed renewable energy, residential adoption faces a massive friction point: site assessments.
Traditional solar and rainwater feasibility evaluations require manual engineering site surveys, expensive LiDAR drone scanning, and opaque hardware estimations. In many developing and peri-urban regions, this barrier excludes everyday homeowners, local businesses, and communities from transitioning to clean power.
We built Skyield to eliminate this friction. Our vision was to give anyone, anywhere on Earth the ability to drop a pin on their rooftop and receive an instant, mathematically rigorous microclimate, solar photovoltaic, and rainwater harvesting feasibility assessment in seconds—for free.
What it does
Skyield acts as an all-in-one property-level climate feasibility engine:
- Interactive Geospatial Pinpointing: Pinpoint any rooftop globally with Google Satellite Hybrid layers and dynamic viewport tile caching—streaming high-resolution tiles on demand and pruning off-screen cache to support low-bandwidth devices.
- Real-Time Atmospheric Telemetry: Fetches 7-day Global Horizontal Irradiance (GHI) in $\text{MJ/m}^2$, direct beam radiation, daylight duration, and precipitation metrics from Open-Meteo with scientific solar climate regression fallback.
- AI Rooftop Vision & Shading Analysis: Leverages multimodal computer vision to classify rooftop materials (e.g., galvanized metal, clay tile), determine usable surface area, detect localized shading obstructions (HVAC, tree canopies), and estimate optimal installation tilt angles $\theta$.
- Instant 60fps Hardware & Sizing Simulation: Computes photovoltaic DC capacity, Inverter Loading Ratio (ILR), battery reserve storage, annual utility cost savings, break-even payback timelines, and EPA-standard carbon offsets in real time.
- One-Click Feasibility Dossier Export: Generates clean, printer-friendly one-page technical reports for homeowners, solar contractors, and policy planners.
Mathematical & Physical Foundations
Skyield computes clean energy and water yields using rigorous physics-based equations:
Photovoltaic Daily & Annual Energy Yield: $$E_{\text{daily}} = A_{\text{usable}} \times \text{GHI}{\text{avg}} \times \eta{\text{panel}} \times \text{PR} \times (1 - S_f)$$ $$E_{\text{annual}} = E_{\text{daily}} \times 365$$ (Where $A_{\text{usable}}$ is usable roof area, $\text{GHI}{\text{avg}}$ is solar irradiance, $\eta{\text{panel}} \approx 20.0\%$ dynamic module efficiency, $\text{PR} = 0.75$ system performance ratio, and $S_f$ is the AI-detected shading factor).
Inverter Sizing Optimization (ILR $\approx 0.90$): $$P_{\text{inv}} = \max\left(1.0\text{ kW}, \text{Round}\left(P_{\text{DC, total}} \times 0.90\right)\right)$$
Annual Rainwater Harvesting Capture: $$V_{\text{rain}} = A_{\text{usable}} \times P_{\text{annual}} \times C_r$$ (Where $P_{\text{annual}}$ is precipitation in $\text{mm}$ and $C_r = 0.85$ is the smooth roof runoff coefficient).
Carbon Offset & Tree Equivalence: $$\text{CO}2\text{ Offset (kg/yr)} = E{\text{annual}} \times 0.42\text{ kg CO}_2/\text{kWh}$$ $$\text{Trees Equivalent} = \frac{\text{CO}_2\text{ Offset}}{21.77\text{ kg CO}_2/\text{tree/yr}}$$
How we built it
- Frontend Application: Built with React, Vite, and Tailwind CSS v4 utilizing an accessible, responsive glassmorphic design system.
- Geospatial & Mapping: Implemented with Leaflet and React-Leaflet, integrating Google Maps Hybrid Satellite layers with custom viewport memory management.
- Atmospheric Telemetry: Direct browser integration with Open-Meteo Forecast APIs paired with an autonomous latitude-based atmospheric regression engine for guaranteed offline/fallback uptime.
- AI Vision Engine: Powered by Google Gemini Multimodal Vision API for rooftop geometry classification, roof surface condition assessment, and obstruction detection.
- Hosting & Infrastructure: Deployed globally with zero-latency edge delivery via Vercel.
Challenges we ran into
- Browser Geolocation Reliability on Hardware Without GPS:
- Problem: When testing on desktop browsers without hardware GPS chips, the standard Geolocation API returned
POSITION_UNAVAILABLE(Code 2). - Solution: Implemented intelligent fallback handling that differentiates error codes and gracefully falls back to IP-based approximate coordinate resolution (
ipapi.co) when device GPS is unavailable.
- Problem: When testing on desktop browsers without hardware GPS chips, the standard Geolocation API returned
- 60fps Real-Time Slider Reactivity:
- Problem: Recalculating hardware kits, battery sizing, and financial curves using asynchronous backend roundtrips caused lag during rapid slider dragging.
- Solution: Engineered a synchronous client-side physics computation engine wrapped in React
useMemo, allowing instant 60fps re-renders across all charts and hardware gauges without network latency.
- Print & PDF Dossier Formatting:
- Problem: Default browser printing captured map controls, zoom bars, and dark glass backgrounds, leading to unreadable prints and ink waste.
- Solution: Designed a dedicated
@media printlayout that automatically strips interactive map canvases and renders a clean, high-contrast executive property feasibility summary.
Accomplishments that we're proud of
- Zero-Latency Simulation: Achieved instantaneous 60fps recalculations across all hardware specifications and solar charts during continuous slider adjustments.
- Resilient Multi-Tiered Architecture: Engineered an application that operates seamlessly even when third-party telemetry services are throttled, utilizing direct client-side meteorological APIs and latitude regression models.
- Holistic Sustainability Scope: Combined solar photovoltaic modeling with rainwater catchment analysis and EPA-standard ecological equivalencies (trees planted, utility bill savings, and water independence).
- Production-Ready Edge Deployment: Successfully packaged, built, and deployed the full application to Vercel with zero cold starts.
What we learned
- Physics-Grounded AI: Combining AI where it excels (computer vision, roof geometry classification, pattern detection) with deterministic physics and atmospheric equations produces far more accurate, trustworthy, and actionable energy forecasts than unconstrained LLM predictions alone.
- Low-Bandwidth Geospatial Engineering: Dynamic viewport tile caching and memory management are critical for ensuring satellite mapping tools remain fast and responsive in regions with limited network connectivity.
What's next for Skyield
- Automated Local Utility Tariff Sync: Integrating regional electricity rate APIs to calculate dynamic net-metering and feed-in tariff ROI down to the municipal level.
- Multi-Polygon Roof Tracing: Allowing users to draw custom multi-angle roof facets directly on satellite tiles for complex architectural layouts.
- Installer Marketplace Integration: Connecting analyzed homeowners directly with verified local solar contractors and micro-financing institutions.
Built With
- artificial-intelligence
- clean-energy
- climate-tech
- computer-vision
- environmental-science
- fastapi
- gemini-api
- geospatial
- green-tech
- javascript
- leaflet.js
- microclimate
- open-meteo
- physics-engine
- python
- rainwater-harvesting
- react
- renewable-energy
- solar-power
- sustainability
- tailwindcss
- vercel
- vite
- web-app
