Inspiration
Electrical failures such as overloads, overheating, and short circuits are among the leading causes of industrial downtime, equipment damage, and safety hazards. In many environments, monitoring systems are reactive rather than predictive — faults are detected only after significant damage has occurred. We were inspired to build a solution that transforms raw electrical sensor readings into actionable intelligence using artificial intelligence. The goal was to demonstrate how large language models can move beyond chat applications and into real-world industrial safety systems. Smart Power Guardian was created to bridge the gap between electrical monitoring and AI-powered decision support.
What it does
Smart Power Guardian is an AI-powered electrical fault detection and advisory system.
The system:
Analyzes voltage, current, and temperature readings from uploaded CSV files
Calculates power consumption
Detects abnormal conditions such as:
Overload
Short circuit risk
Overheating
Normal operation
Assigns severity classifications
Generates structured, engineering-style safety recommendations using Groq LLaMA 3.3 70B
Displays results in an industrial dashboard format with graphs and tables
It converts raw sensor data into real-time diagnostic intelligence.
How we built it
The system was built using a modular AI pipeline:
Data Processing Layer
Python and Pandas for sensor data handling
Rule-based detection logic for fault classification
AI Reasoning Layer
Groq API for high-speed LLaMA 3.3 70B inference
Structured prompts to generate engineering-grade recommendations
Frontend & Deployment
Gradio for interactive UI
Hugging Face Spaces for cloud hosting
HTML, CSS, Bootstrap, and JavaScript for product landing page
Google AI Studio (Gemini) for prompt refinement and UI structuring
Challenges we ran into
Designing reliable fault detection logic Creating threshold-based logic that mimics real industrial scenarios required careful tuning to avoid false positives.
Integrating LLM output into structured engineering responses Large language models tend to generate conversational text. We had to refine prompts to ensure structured, technical recommendations.
Ensuring smooth deployment Configuring API keys securely and deploying to Hugging Face Spaces required debugging environment variables and dependency management.
Designing a professional dashboard Balancing clean UI design with industrial aesthetics was a challenge, especially while keeping everything responsive.
Accomplishments that we're proud of
Successfully integrated Groq LLaMA 3.3 70B into a functional industrial safety prototype
Built a full-stack working application with live deployment
Created a professional product website and demo video
Demonstrated real-world application of AI in electrical engineering
Designed a scalable architecture suitable for IoT and smart factory expansion
Most importantly, we moved AI beyond theory and into a practical industrial safety use case.
What we learned
Large language models can be adapted for structured technical domains with proper prompt engineering
Industrial applications require clarity, precision, and reliability — not just AI creativity
Deployment and system integration are as important as model performance
Clear UI/UX significantly improves perceived system intelligence
We also learned how to structure AI systems in a modular way for scalability.
What's next for Smart Power Guardian – AI Fault Detection System
The current version uses simulated CSV data. Future development includes:
Real-time IoT sensor integration
Edge deployment on microcontrollers
Predictive maintenance modeling using time-series forecasting
Real-time streaming dashboards
Cloud-scale analytics for industrial plants
Automated alert systems via SMS or email
The long-term vision is to evolve Smart Power Guardian into a full-scale industrial AI monitoring platform for smart factories and energy systems.
Log in or sign up for Devpost to join the conversation.