LambdaEdge AI

Serverless AI that transforms industrial telemetry into actionable insights using AWS Lambda and DynamoDB.

Comment 1

LambdaEdge AI – Predictive Fault Diagnostics for Industrial Telemetry

Inspiration

With over a decade of hands-on experience in instrumentation and control systems, both in industrial operations and academic research, I’ve seen how unplanned equipment failures can lead to costly downtime and safety risks. Some industries that include small and medium-sized enterprises (SMEs) still rely on legacy systems that offer little more than basic monitoring, with no predictive capability.

Through my recent research, I became deeply interested in bridging the gap between traditional systems and modern, intelligent diagnostics. I wanted to demonstrate that predictive insights don’t require expensive overhauls, and that cloud-native telemetry and diagnostics can be both simple and scalable, especially with AWS services. I’m also mindful of the cyber-physical security concerns in this space and recognize the proactive role developers and cloud providers are playing to safeguard critical operations.

It is this curiosity that led to the development of LambdaEdge AI, which is a fully autonomous fault diagnostic system built on AWS Lambda, DynamoDB, and an interactive Streamlit dashboard. It features a lightweight, cloud-native solution that mimics real industrial telemetry, processes it intelligently, and provides early warnings of potential failures, using scalable serverless tools.

What it does

LambdaEdge AI is a real-time predictive fault diagnostics system that:

  • Simulates telemetry (temperature, vibration, status) from 3 virtual industrial devices
  • Detects abnormal patterns or fault signatures using a scheduled AWS Lambda process
  • Logs all telemetry and fault events to a DynamoDB time-series store
  • Visualizes system health and diagnostics on a public Streamlit dashboard
  • Secures cloud operations using scoped IAM roles and scheduled automation

In short, it mimics a smart edge system that thinks proactively before equipment failure occurs.

How the Interface Works

The public access dashboard is hosted on Streamlit Community Cloud and has the following characteristics:

  • A sidebar to select the Data Range* (in minute) with option for **Device and Sensor to view.
  • Access summary metrics for each device (average readings, uptime, fault count)
  • View Live Telemetry sensor data in a downloadable CSV format.
  • Real-Time Telemetry in Gauge and Chart: color-coded device status: Green (Normal), Yellow (Warning), Red (Fault)
  • Predictive Diagnostic for ML-inferred anomaly risks and failure probabilities.

The dashboard provides an intuitive experience. Just like a SCADA-lite interface fused with AI insights, accessible in any browser.

How I Built It

I developed and deployed the entire system end-to-end, from backend logic to UI, using the following stack:

  • AWS Lambda:
    • Lambda Injector: Simulates telemetry for 3 virtual devices with realistic behavior and random fault injection.
    • Lambda Edge Handler: Processes incoming data, performs diagnostic analysis, and tags fault events.
  • EventBridge: Triggers the injector Lambda function every 60 seconds.
  • DynamoDB: A time-series table (FaultEventLog) stores all telemetry and fault classification data.
  • Streamlit: A front-end dashboard that visualizes real-time device data, system status, and diagnostic insights.
  • IAM: Scoped permissions ensure secure Lambda execution and controlled access to AWS services.

Challenges I Faced

  • Designing realistic fault models using only synthetic data
  • Managing Lambda and EventBridge scheduling concurrency
  • Ensuring real-time dashboard updates without overloading compute
  • Tuning DynamoDB structure for time-series performance
  • Building an interface that’s not only functional, but clear and insightful

Accomplishments I'm Proud Of

  • Built a complete predictive diagnostics system solo in a constrained timeframe
  • Delivered real-time insights via a publicly accessible cloud dashboard
  • Achieved live telemetry simulation, detection, and visualization, all serverless
  • Created a system that can scale to hundreds of edge devices with minimal cost

What I Learned

  • How to orchestrate serverless automation using Lambda + EventBridge + IAM
  • Advanced Streamlit dashboarding techniques for real-time display
  • Building resilient, modular systems for predictive maintenance use cases
  • Applying AI logic to fault classification and edge simulation

What’s Next for LambdaEdge AI

  • Integrate machine learning models (e.g., with Amazon SageMaker or TinyML in Lambda)
  • Support real industrial protocols like MQTT, OPC-UA, or Modbus
  • Extend the dashboard with MongoDB Charts or Grafana cloud integration
  • Add authentication and role-based views for different user types (engineers, managers)
  • Package as a reusable template for manufacturers, plant operators, and AI startups

Powered by AWS. Designed to save time, costs, and equipment, one smart diagnostic at a time.

Built With

Share this project:

Updates