CampusWatch is an interactive 3D energy monitor for a university campus. It finds electricity wasted in individual rooms, such as an AC, fan or lights left on after everyone has left, or a sudden abnormal power spike, and alerts the person responsible.

On the 3D model of our campus, you hover Kashyapa Bhavanam, AB1 or Maithreyi Bhavanam, pick a block, click a floor and see every room (like C305) colored green, amber or red by its live power against what is normal for its occupants. A rule-based detector compares each room’s consumption with its occupancy and its own usual pattern. When something is wrong, an alert goes to the room’s occupant and escalates to the warden if ignored, while wasted kWh is tracked for every room, floor, block and building.

An open-weight language model (Qwen2.5-7B, running locally through Ollama) writes the alerts in plain language, explains why a room was flagged and answers questions like “Which block wasted the most power today?” It only receives numbers computed by the detector, and it works offline, so room-level data stays on campus.

The prototype runs on synthetic data in the same format real room sensors would send, such as power meters and presence sensors. Moving to live data means changing only the data source. The project is open source on GitHub.

One-line version: CampusWatch is an open-source 3D campus monitor that detects electricity wasted in empty rooms and alerts the responsible person, using an offline open-weight AI to explain every alert.