Inspiration Electricity prices change every 30 minutes, but existing tools force users to input rigid deadlines. Real schedules are fuzzy ("I leave around 7:30, maybe 8:15"). We built Tempo to manage both grid volatility and human ambiguity, putting a precise financial price on schedule flexibility.

What it does Tempo is an uncertainty-aware energy scheduler. Users type their messy routine into a text box. Tempo extracts three scenarios: Earliest, Expected, and Latest deadlines. It then runs a Monte Carlo simulation and constraint solver to show the expected cost, carbon footprint, and the "Flexibility Spread": the exact money saved by remaining flexible.

How we built it The backend uses FastAPI, with a React/Vite frontend.

Grid Engine: Ingests Octopus Agile and UK Carbon data, using quantile regression and Monte Carlo simulations for price forecasting.
Human Engine: Uses the Gemini API with strict Structured Outputs (temperature=0.0) to parse semantic schedules.
Solver: Google OR-Tools mathematically guarantees the cheapest optimal running slots.

Challenges we ran into The UK Daylight Saving Time transition happened mid-hackathon (25 October 2026), creating a 25-hour night. We built a strict DST-aware layer to safely translate Gemini’s local wall-clock times into UTC. We also built a deterministic Python validator to instantly reject contradictory LLM outputs rather than silently executing a flawed schedule.

Accomplishments that we're proud of We successfully isolated LLM semantic parsing from deterministic math. We’re also proud of our Savings Ledger, which back-tests our algorithm against real London smart-meter profiles, and our Herd Simulator, which models how staggering start times prevents rebound grid peaks.

What we learned Large Language Models shouldn't generate statistical probabilities. By forcing Gemini to extract discrete time scenarios and leaving the probability math to the Monte Carlo engine, we built a much safer, trustworthy application.

What's next for Tempo Integrating routine learning from historical smart-meter data to anticipate flexibility needs, expanding to business EV fleets, and adding smart plug integration for automated control.