Volt-Wise

Inspiration

With utility providers like Xcel Energy transitioning millions of residential customers to Time-of-Use (TOU) pricing and introducing strict "Demand Charges," we noticed a massive gap in how ordinary people interact with their power bills. Most people only find out they ran their dryer during a peak hour when they receive a surprisingly high bill at the end of the month. We wanted to build a proactive tool that doesn't just show you past data, but actively acts as a financial planner for your home's energy consumption.

What it does

Volt-Wise is an intelligent, localized smart-home energy planner. It allows users to register their household appliances (like EVs, ACs, and Electric Ovens) and schedule when they want to run them. The application cross-references these schedules with live Time-of-Use grid data and calculated concurrent kW draw to provide real-time cost projections. If a user tries to run a heavy appliance during "Peak Pricing," our custom deterministic AI Savings Advisor instantly flags the action, calculates the exact dollar amount they will be penalized, and suggests an optimal off-peak time to shift the load.

How we built it

We architected Volt-Wise as a full-stack web application.

• Frontend: We used Next.js 14 and React to build a highly responsive, glassmorphic UI styled with Tailwind CSS.
• Backend: We created a high-performance Python FastAPI server to handle the heavy mathematical lifting.
• Data & Auth: We used an SQLite database with SQLAlchemy for persistent storage, and implemented JWT token authentication from scratch to secure user data.
• Core Logic: Instead of relying on a slow, generic LLM, we engineered a custom, zero-latency deterministic algorithmic engine in Python that perfectly models complex TOU rate tiers and demand charge thresholds.
• AI Pair Programming: We built the entire stack alongside Google Antigravity, utilizing it as our primary AI developer and pair-programmer to rapidly iterate on both the frontend design and complex backend state management.

Challenges we ran into

One of our biggest hurdles was accurately modeling Demand Charges. Unlike flat kWh rates, a demand charge penalizes the user based on the single highest spike of concurrent kilowatts pulled within an hour. We had to write complex local algorithms to simulate overlapping appliance wattages and determine if scheduling a new device would accidentally push the user's home over their monthly kW threshold, which required very precise floating-point math across the frontend UI and the Python backend. Another challenge was timezone normalization—ensuring that a user scheduling a device at "5:00 PM" in their local browser strictly aligned with the utility provider's strict peak-hour UTC timestamps.

What we learned

We learned a tremendous amount about the intricacies of modern power grids. Specifically, we learned how utility companies structure flat energy costs versus peak-demand penalties, and how much money consumers can save just by shifting their appliance usage by a few hours. Technically, we leveled up our ability to integrate Next.js Server Components with a secure, JWT-authenticated Python FastAPI backend, and we learned how to design deterministic AI rule-engines that provide instant, actionable financial advice without relying on external generative models.

Built With

• and-jwt-authentication
• and-typescript
• css
• fastapi
• next.js
• powered-by-a-python-fastapi-backend-using-sqlalchemy
• pydantic
• react
• sql
• sqlite
• tailwind
• tailwind-css
• typescript