Back to all updates

5 months ago

Congratulations to the Winners of the AWS AI Agent Global Hackathon!

Hi there,

For those who may have missed the big reveal at re:Invent, we are thrilled to celebrate the winners of the AWS AI Agent Global Hackathon 🎉!

The quality and innovation demonstrated across all submissions were truly exceptional, showcasing the amazing potential of AI agents built on AWS 🤖. We saw a stunning display of skill, creativity, and dedication. A massive thank you to everyone who devoted their time and effort to building and submitting a project. 

 

A huge round of applause for our overall top winners 🏅

  • 1st Place: EcoLafaek

    • EcoLafaek addresses critical waste management challenges in Timor-Leste by combining a citizen-led mobile reporting app with a sophisticated autonomous AI agent powered by Amazon Bedrock Nova-Pro. This full-stack solution impressively leverages multi-modal reasoning and AgentCore to autonomously chain tools, classifying waste images and generating real-time data visualizations to map pollution hotspots. By providing authorities with actionable insights and deep semantic search capabilities, this project effectively empowers the community to tackle the over 100 tons of daily uncollected waste in Dili.

  • 2nd Place: AegisAgent

    • AegisAgent transforms manual insurance claim reviews into an automated, explainable process by utilizing AWS Kiro to orchestrate specialized agents for evidence curation, policy interpretation, and compliance reasoning. Remarkably, this full-stack proof of concept was developed in just five days with 100% of the code generated by AI, demonstrating the immense potential of AWS Bedrock models and semantic indexing in handling complex data. The resulting system delivers a transparent, defensible coverage decision workflow that effectively debates and resolves ambiguities in claim artifacts and policy documents.

  • 3rd Place: Province

    • Province transforms the complex tax filing process into a seamless natural conversation by utilizing a sophisticated multi-agent architecture to extract data, explain calculations, and auto-fill IRS forms with CPA-level accuracy. The system leverages AWS Bedrock and Claude 3.5 Sonnet to power an innovative autonomous FormMapping pipeline, which uses agentic reasoning to map PDF fields dynamically and achieve 100% accuracy on Form 1040. By combining intelligent document processing with a verified IRS knowledge base, this platform effectively removes the barriers of expensive professional help and confusing DIY software.

Best of Category Winners ⭐

 

Congratulations once again to all the winners! Your projects truly stand out.

And to every participant: thank you for being a part of this challenge. We encourage you to continue developing your projects and exploring the power of AI agents on AWS.