Wattora: An ML-Powered Solar Forecasting and Energy Optimization Platform

A Full-stack app that forecasts solar generation and household electricity demand with Random Forest models, then turns those predictions into energy-use recommendations. Covers the architecture, tech stack, ML pipeline, secure authentication, and a plan for deploying it on AWS.

Pooja Ramniwas

Published Oct 11, 2026

Last modified Oct 11, 2026

Solar panels are easy to install but hard to use well. The sun peaks at midday, while most homes use the most power in the morning and evening. Owners usually see only what already happened, so they still buy grid power when their own solar could have covered it.
Wattora closes that gap. It forecasts how much solar power a site will produce and how much electricity it will need, compares the two, and says what to do about it.
Source code:  https://github.com/Pikanshi1/Wattora 
Problems it solvesUnpredictable solar output. It forecasts generation through the day, scaled to the site’s installed capacity.
Guesswork about demand. It forecasts hourly demand, personalised from the site’s daily usage and enabled devices.
Mismatch between the two. The optimizer reports surplus, grid power needed, and self-consumption (the share of generated solar the home actually uses).
Numbers without advice. It gives a plain-language recommendation based on the user’s priority (solar, savings, or balanced) and battery reserve.
Who it is forHomeowners with rooftop solar, and households considering it
Small businesses and facility managers running several sites
Students and researchers who want a complete end-to-end ML project
Developers who want to contribute to an open source project
Where we are todayThe backend is complete: accounts, sites, devices, both forecasts, and the optimizer. The dashboard pages are built. Forecasts currently use illustrative weather values, and sign-in is not yet connected to the backend. Those are our first next steps.
Future impactBetter timing of flexible loads can mean lower bills, more of the clean energy actually used, and less pressure on the grid at peak hours. It also gives students and small businesses access to tools usually built for utilities. We have not yet measured savings in real homes, and results will vary by site and tariff.
Features we plan to addLive weather integration and smart-meter data
Forecast accuracy tracking and confidence ranges
Appliance scheduling for flexible devices
Real cost and savings estimates, with time-of-use tariffs
Battery and EV-charging planning
Email and push alerts, and a mobile-friendly app
Carbon-footprint view and multi-language support
Automated model retraining and explainable predictions
AWS deployment: S3 and CloudFront for the frontend, Fargate or App Runner for the API, RDS for PostgreSQL, SES for email, and Secrets Manager

# solar # billing # cost-optimization # amazon-ec2

Builder Center

Wattora: An ML-Powered Solar Forecasting and Energy Optimization Platform

Pooja Ramniwas

Problems it solves

Who it is for

Where we are today

Future impact

Features we plan to add