GridNinja

How We Built It

We developed a physics-informed graph neural network that integrates real-world grid dynamics, energy generation profiles, and data center load behavior. This hybrid approach enables accurate simulation and optimization across multiple operational modes.

Challenges We Ran Into

The scope of the problem quickly expanded—there were many additional features and modeling layers we wanted to implement. Time constraints ultimately limited how far we could push the system in this iteration.

Accomplishments We’re Proud Of

We successfully integrated learning-based models with real-world data to simulate and optimize multiple efficiency regimes. The result is a flexible digital twin capable of capturing complex interactions between power grids, data centers, and carbon intensity.

What We Learned

We gained hands-on experience implementing secure authentication with Auth0, building APIs to connect frontend and backend systems, and deploying learning models within a full-stack application.