SmartOps

SmartOps is an AI-powered ticket triage system that automatically analyzes, categorizes, prioritizes, and routes support tickets to the right teams, reducing response time and improving efficiency.

Comment

Inspiration

Support teams receive hundreds of customer emails daily about payments, deliveries, and product issues. Manually reading and categorizing each ticket is time-consuming and error-prone. I wanted to build an AI system that could instantly understand customer issues and route them to the right team - saving time and improving response speed.

What it does

SmartOps automatically processes customer support tickets using AI:

  • Receives tickets via email or bulk upload (CSV/JSON)
  • Analyzes content using Amazon Bedrock Nova Lite
  • Categorizes into 7 types (payment, delivery, product quality, etc.)
  • Assigns severity (Critical, High, Medium, Low)
  • Routes to appropriate teams
  • Sends email alerts for critical issues
  • Displays everything in an interactive dashboard

Key feature: Send an email to support@jobsaddaa.in and it's automatically analyzed and categorized in 30 seconds - zero manual work!

How we built it

Technology Stack:

  • Amazon Bedrock Nova Lite v1 for AI analysis
  • AWS Lambda (Python 3.13) for serverless processing
  • S3 for storage and event triggers
  • DynamoDB for database with fast queries
  • SES for email receiving
  • SNS for critical alerts
  • Streamlit for dashboard
  • SAM for infrastructure as code

Architecture: Email/Upload → S3 → Lambda → Bedrock AI → DynamoDB → Dashboard (with SNS alerts for critical tickets)

The system is fully serverless, event-driven, and scales automatically.

Challenges we ran into

  1. Email Parsing: Raw emails have complex formats. Solved using Python's email library to cleanly extract subject, sender, and body.

  2. AI Response Validation: Bedrock sometimes returns invalid JSON. Built retry logic with validation and fallback defaults.

  3. SES Configuration: Email receiving requires MX records and receipt rules. Created detailed setup guide and automated with SAM.

  4. Cost Optimization: Used Nova Lite (most cost-effective), optimized prompts, and set token limits to keep costs under $1 per 100 tickets.

Accomplishments that we're proud of

  • Fully functional production system with live dashboard
  • Email-to-ticket automation working end-to-end
  • 90%+ AI accuracy in categorization
  • Cost-effective at scale (less than $1 per 100 tickets)
  • Clean, well-documented code with error handling
  • Comprehensive documentation and security audit

What we learned

Technical: Amazon Bedrock integration, serverless architecture, event-driven design, email processing with SES, NoSQL database design with DynamoDB, and Infrastructure as Code with SAM.

AWS Ecosystem: How different AWS services work together, IAM permissions, CloudWatch debugging, and cost optimization strategies.

Soft Skills: Breaking complex problems into smaller pieces, writing clear documentation, and designing intuitive user experiences.

Share this project:

Updates