Arkos
AI-powered energy analytics and advisory platform
Project metadata
Overview
An energy analytics and advisory platform for households and small businesses, pairing LSTM consumption forecasting with a RAG chatbot grounded in the user's own energy bills and reports.
Problem
Energy inefficiency disproportionately affects low-income households and small businesses that lack access to sophisticated energy management tools. Rising energy costs and environmental concerns increase the need for an accessible, AI-powered alternative.
Features
Smart energy forecasting
- LSTM neural networks for consumption prediction
- Pattern recognition for usage trends and anomalies
- Multi-horizon forecasting, daily through monthly
Intelligent document analysis
- RAG-powered chatbot that reads energy bills and reports
- PDF processing with automatic data extraction
- Contextual insights based on the user's specific usage patterns
Interactive analytics dashboard
- Real-time visualizations via Chart.js
- Cost breakdown analysis with recommendations
- Comparative analytics to track improvement over time
Cost optimization engine
- Personalized recommendations from usage patterns
- Peak-hour analysis to minimize high-cost consumption
- Savings calculator
Key algorithms and models
LSTM energy forecasting
RAG document analysis
Full stack
React 19 · TypeScript · TailwindCSS 4.1 · Vite · Chart.js · React Router · Axios · Lucide React · Motion · Flask · Flask-CORS · Python · TensorFlow · Keras · Gemini API · ChromaDB · PyPDF · PyMuPDF · Camelot · Tiktoken · Pandas · NumPy · scikit-learn
Results
LSTM architecture: multi-layer with dropout | Yes | |
RAG: 800-token overlapping chunks, hybrid retrieval | Yes | |
ChromaDB + Gemini embeddings | Yes |