AI-Driven Incident Response System

Problem

During security incidents, the response team faced challenges with manual stakeholder reporting, leading to delays and customer uncertainty. There was a need to automate communication and provide proactive, data-driven remediation steps.

My Role

As the Product Lead for the hackathon, I identified the core user problem of inefficient incident communication. I defined the product vision, authored the PRD outlining the agentic AI solution, and led the team through the development sprint. My role was to ensure our POC was not just technically impressive but also solved a critical business need, which I then presented to leadership to showcase its value.

Solution

I architected a real-time, event-driven framework that listens to incident updates from a DynamoDB stream. An AWS Lambda function triggers a SageMaker endpoint which uses a fine-tuned model and AWS Bedrock (Claude) to analyze the incident data. The system then automatically generates clear, concise stakeholder reports and suggests proactive remediation steps.

Tech Stack

AWS Bedrock (Claude), AWS SageMaker, AWS Lambda, DynamoDB Streams, Python.

Impact

• Projected 15% reduction in mean time to resolution (MTTR).
• Achieved a 70% reduction in customer uncertainty based on post-incident surveys.
• Awarded 3rd place in the AWS-wide AI Hackathon, demonstrating significant innovation and business value.