Inspiration

In large-scale distributed systems, logs often overflow with unprioritized noise. We wanted to create a smart, serverless alert system that filters, classifies, and responds to incidents in real time — all built around AWS Lambda.

What it does

The AI Incident Alert System listens to errors from AWS CloudWatch Logs, classifies them into Critical, High, Medium, and Low severities using Lambda logic, and stores them in DynamoDB. If a Critical issue is detected, a Lambda function triggers an immediate email alert via Amazon SES.

A frontend dashboard (React) powered by API Gateway + Lambda displays all incidents with filtering by severity and status (open, resolved), helping teams respond faster and smarter.

How we built it

  • AWS Lambda: Core logic for error parsing, classification, email alerts, and API backend.
  • Amazon CloudWatch Logs: Source of error logs from subscribed systems.
  • Amazon DynamoDB: Storage for incidents with structured fields.
  • Amazon SES: Sends alert emails on Critical errors.
  • API Gateway: Exposes Lambda functions to the frontend.
  • ReactJS Dashboard: Provides an interactive UI to view and filter incidents.
  • AWS SAM: Used for packaging and deploying infrastructure. ## Challenges we ran into
  • Handling varied log formats across services.
  • Dealing with SES sandbox email restrictions.
  • Maintaining low latency in a fully serverless pipeline.
  • Designing a clear yet powerful UI for incident management.

Accomplishments that we're proud of

  • Fully serverless architecture using only AWS Lambda for all compute needs.
  • Real-time classification and alerting pipeline.
  • Scalable design that can plug into any system emitting CloudWatch logs.
  • A functional and clean dashboard to monitor incidents in real time.

What we learned

  • Deep dive into Lambda’s integration capabilities across AWS services.
  • Practical use of SES, DynamoDB, and API Gateway in real-world scenarios.
  • Serverless design patterns and error handling at scale.
  • Managing IAM roles and security while maintaining developer velocity.

What's next for AI Incident Alert System

  • Slack integration for feedback (“Resolved”, “Noise”, “Critical”) to retrain model.
  • Web dashboard to view active alerts & ML score.
  • Auto-ticket creation in Jira/ServiceNow via Lambda.
  • ML-based anomaly & severity classification
  • AI-driven deduplication & grouping.
  • Slack, SMS, Webhooks, team-based routing.

Built With

Share this project:

Updates