Ebrahim Khiyami

Gen-AI Assisted AWS Well-Architected Review

[Deliverable: re:Invent2024 session, customer case study, AWS Sample Solution and more].

Short Summary

This is WA-Gen-AI app. My second Gen-AI POC that I built this year after the success I had in building the first one, the YouTube Summarizer.

In this POC, I use Gen-AI to streamline the Well-Architected Review for an application written in Infrastructure as Code (IaC), e.g., CloudFormation, Terraform, or else.

This solutions received a lot of attractrions from my collegues and from AWS customers. After I published I was reached out by an awesome group of SAs in my team who took it and improve it and went through the necessary reviews to make it an offical solutiosn on AWS Repo.

wa-gen-ai

In recent weeks, Commonwealth Bank, an Australian multinational bank, used this solution to streamline their Well-Architected Reviews.

They were so impressed that they wanted to showcase it publicly on stage at re:Invent 2024, AWS’s largest annual conference.

You can try my personal version here. Check my blog post on AWS Community here

This is the the official version from AWS Sample.

You can watch the demonstation on re:Invent stage here.

Long Summary

If you’ve been an AWS customer for some time you should be familiar with the Well-Architected Framework . It’s the ultimate collection of AWS best practices across 6 pillars (Operational Excellence, Security, Reliability, Performance Efficiency, Cost Optimization, and Sustainability).

and the Well-Architected Review is the process to validate your architecture for these best practices. The results would be a list of identified risks (high, medium) in the architecture and action items to improve them.

The WA-Gen-AI App takes a CloudFormation template uploaded by the user, uses AWS Bedrock for Gen-AI and Streamlit for the front-end, and Python to interact with the AWS Well-Architected Tool, iterate through best practices, and complete the review.

In my testing, I was able to complete the review in a few minutes, compared with a few hours that were traditionally needed by customers or the AWS team to bring everyone into the same room and answer the 57 of review questions manually.

In the 3-tier classic application I tested, the WA-Gen-AI app was able to answer about 50% of the questions and save hours of manual review.

More details

wa-gen-ai-post

wa-gen-ai-post