-
-
Triagelytics
Inspiration
Modern systems generate massive amounts of logs, but teams often struggle to extract meaningful insights quickly. We wanted a simple, automated way to analyze logs, categorize issues, and surface clear action items—without manual digging. That led us to build Triagelytics: an AI-driven, serverless log triage system powered by AWS and Bedrock LLMs.
What it does
Triagelytics securely analyzes uploaded logs from an S3 bucket, processes them through Lambda functions, runs intelligence analysis using Bedrock LLMs, and then triages incidents with prioritized action items. Users sign in via Cognito and get a clean React dashboard to explore findings, mark resolutions, and track remediation steps.
How we built it
We built Triagelytics on a fully serverless architecture using AWS services end to end.
Logs are uploaded to S3.
Lambda functions handle extraction, parsing, and LLM-based analysis in Bedrock.
A triage engine categorizes and scores incidents.
Authentication is powered by Cognito.
The frontend is a React app integrated with the backend API.
Infrastructure automation is done using AWS SAM CLI and Terraform, along with deployment and cleaning scripts for repeatable environments.
Development flows and orchestration leverage KIRO for structure and efficiency.
Challenges we ran into
Setting up the LLM prompts and tuning them for consistent log analysis was challenging. Managing serverless state transitions, IAM permissions, and cross-service communication required careful configuration. We also had to ensure that triage outputs were reliable and that the React dashboard displayed results clearly and securely.
Accomplishments that we're proud of
We successfully built an end-to-end system that takes raw logs and produces actionable insights automatically. The seamless integration of Bedrock LLMs with Lambda, the clean triage flow, and our automated deployment pipeline are major wins. We’re also proud of how intuitive and fast the React dashboard feels.
What we learned
We learned how to craft more structured prompts for LLM-based log analysis, tune Bedrock for consistency, and design better serverless workflows. We also deepened our understanding of Terraform, SAM, and Cognito integration, and how to build a scalable, secure pipeline with minimal operational overhead.
What's next for Triagelytics
We plan to expand support for more log formats, add real-time streaming analysis, integrate alerting through Slack and email, and improve visualization dashboards. We also want to introduce a recommendation engine that suggests fixes automatically and add multi-tenant support for enterprise deployment.
Log in or sign up for Devpost to join the conversation.