WattMizan

Inspiration

Small businesses face rising energy costs but often lack the time, data, or budget for a first-pass energy audit. With my Energy Engineering background and firsthand knowledge of resource-constrained businesses in Jordan, I wanted to make the first useful step take minutes instead of weeks.

What it does

WattMizan asks for a business type, monthly electricity bill, electricity rate, and daily operating hours. It instantly provides a transparent screening estimate of annual savings, energy addressed, and avoided emissions. Gemini then turns the submitted context into three prioritized practical actions and one verification question.

WattMizan is a hardware-free screening and decision-support tool. It does not claim to provide real-time monitoring or replace a professional site audit.

How I built it

I built WattMizan as a responsive web application using React, TypeScript, and vinext. A deterministic calculation layer keeps the initial estimate transparent and explainable.

The server calls Gemini 3.5 Flash-Lite through the Gemini API, using an API key connected to a Google Cloud project. Gemini receives validated business inputs and generates a concise action plan grounded in those inputs. The API key remains server-side and is not exposed in the public repository.

The working application is deployed publicly, while the complete source code and setup instructions are available on GitHub.

Challenges

The main challenge was providing useful guidance from very limited information without presenting false precision. I separated the transparent engineering calculation from the AI recommendations, capped the screening factor, constrained the prompt, and added clear verification and safety guidance.

I also added input validation, request-size limits, basic rate limiting, and restrictions that prevent the model from recommending hazardous hands-on electrical, gas, or refrigerant work.

Accomplishments

Built a working no-signup energy screening experience
Produced explainable estimates that update instantly
Connected a live Gemini workflow for practical recommendations
Protected the Gemini API key on the server
Created a focused product that can be tested directly with small businesses
Published the source code and working demo

What I learned

AI is most credible in this use case when it explains and prioritizes transparent calculations rather than inventing precise numbers. Keeping the product narrow also made it possible to build, test, and launch a useful working product as a solo developer.

What's next

The next step is to onboard pilot businesses, add Jordan-specific tariff profiles, gather consented feedback, and compare recommendations against follow-up electricity readings. Later versions could add bill-image extraction and recurring monitoring after the screening model is validated with real users.