GridSense

Inspiration

Energy bills are often opaque and difficult to interpret. Most households lack clear, personalized insight into how daily behaviors directly impact both their monthly costs and environmental footprint. We were inspired to build GridSense to bridge that gap transforming raw energy inputs into meaningful analytics, actionable recommendations, and motivating, gamified feedback that empowers smarter energy decisions.

What It Does

GridSense is a full-stack energy analytics platform that estimates household energy consumption, predicts monthly utility bills using provider-specific rates, and delivers personalized sustainability insights. Users receive:

A predicted monthly energy cost
A detailed carbon footprint breakdown
A dynamic eco-score
Smart, behavior-driven recommendations
Leaderboard benchmarking against similar households By combining predictive modeling with gamification, GridSense makes energy optimization measurable, intuitive, and engaging.

How We Built It

The frontend was developed using React 18, TypeScript, Vite, and Tailwind CSS, enabling a responsive, component-driven dashboard that dynamically renders analytics in real time. The backend is powered by Flask and modular Python services, responsible for:

Input validation and normalization
Utility-aware billing calculations
Carbon footprint computation
Recommendation generation
Eco-score and leaderboard logic
User inputs are enriched through ZIP-based geocoding and weather APIs to incorporate regional and climate-specific variation.
For energy prediction, we trained a Scikit-learn Gradient Boosting Regressor using domain-driven feature engineering.
Hyperparameters were optimized with RandomizedSearchCV, and the model was evaluated using standard regression metrics before being serialized with joblib for production inference within the backend.

This modular architecture ensures clear separation between modeling, analytics logic, and API delivery.

Challenges We Ran Into

One major challenge was designing the system to gracefully handle incomplete or inconsistent user inputs while preventing frontend crashes from missing backend fields. Maintaining strict API contracts between frontend and backend became critical as new features, such as eco-scores and leaderboard metrics, were introduced. Additionally, modeling realistic utility pricing structures, including time-of-use considerations, required careful abstraction to ensure both technical correctness and maintainability.

Accomplishments We’re Proud Of

We successfully built a fully integrated, end-to-end energy analytics platform that transforms structured household inputs into real-time cost, carbon, and engagement insights. We are especially proud of:

Our modular and scalable backend architecture
The integration of provider-specific utility billing logic
The deployment-ready machine learning pipeline
The seamless embedding of gamification features within a clean, responsive user experience

Most importantly, the fact that GridSense feels cohesive, not just a model, not just a dashboard, but a unified product

What We Learned

This project reinforced the importance of:

Strong API design and contract stability
Defensive frontend rendering in data-driven applications
Modular backend architecture for scalability
Translating complex technical metrics into intuitive user experiences

We also deepened our experience in training and validating machine learning models, ensuring they generalize well without overfitting or underfitting, and integrating them into a production-style inference pipeline.

What’s Next for GridSense

Next, we plan to:

Integrate real utility datasets for live pricing accuracy
Enhance the ML-driven recommendation engine with adaptive learning
Add historical usage tracking with authenticated user accounts
Expand beyond California utilities
Deploy GridSense to the cloud to support real-world scalability Our long-term vision is to make GridSense a scalable, data-driven platform that helps households nationwide optimize energy usage while reducing environmental impact