• 

🚀 InfraCopilot AI

Predictive Maintenance & Incident Response for EV Charging Networks

InfraCopilot AI is a full-stack platform that uses machine learning to predict failures in EV charging infrastructure, generate copilot-style maintenance recommendations, and reduce costly downtime.

Built for the Data Pigeon AI Incident Response Challenge at SacHacks.

---

🎯 Problem

EV charging networks face:

• Unexpected failures → downtime
  
• Expensive reactive maintenance
  
• Poor prioritization of repairs
  
• Lost revenue from non-functional chargers
  

Most systems react after failures occur.

---

🧠 Solution

InfraCopilot enables proactive maintenance by:

• Predicting failure probability for each charger
  
• Identifying high-risk units in advance
  
• Explaining why a failure may happen
  
• Recommending actionable fixes
  
• Quantifying potential cost savings
  

⚙️ Tech Stack

Machine Learning

• Python
• scikit-learn
• SMOTE (handling class imbalance)
• Logistic Regression & Random Forest
• Precision-Recall optimization
• Cost-aware threshold tuning

Backend

• FastAPI
• Pandas
• Uvicorn

Frontend

• Next.js (App Router)
• TypeScript
• TailwindCSS

---

📊 Model Performance

• Dataset: 50,000 simulated chargers
  
• Failure rate: ~3%
  
• ROC-AUC: 0.99
  
• PR-AUC: 0.81
  
• Recall (failures caught): 90%+
  
• Precision: ~49%
  
• Cost savings: $300K+

This ensures the model prioritizes catching failures over avoiding false alarms.

---

🔍 Features

📊 Fleet Dashboard

• Total chargers
• Risk distribution (Critical / Warning / Safe)
• Total projected savings

📋 Fleet Table

• Search by charger ID
• Filter by risk level
• Sort by savings or risk
• Pagination (handles 50K+ rows)

🤖 Copilot View

For each charger:

• Failure probability
• Root cause analysis
• Recommended action
• Time-to-failure estimate
• Estimated savings

---

infra-copilot/ ├── backend/ │ ├── main.py │ ├── models_v5/ │ ├── outputs/ │ └── requirements.txt ├── ml/ │ ├── data_generator.py │ ├── train_model.py │ ├── inference_engine.py ├── frontend/ │ ├── app/ │ ├── components/ │ └── lib/ └── README.md

Built With

• fastapi
• next.js
• pandas
• python
• scikit-learn
• smote
• tailwind
• typescript
• uvicorn