Winning the AWS Jam Toronto: An Unexpected Journey

I walked into the AWS office in downtown Toronto last Thursday, April 30th, tired but eager. The plan was simple: grab a coffee, sit in on some training, and hopefully pick up a thing or two about what’s happening in the AWS ecosystem. I certainly didn’t expect this to be a fierce competition, and I absolutely did not expect to leave as part of the winning team.

It’s a very human thing to underestimate what a group of strangers can achieve. When it was time to form teams, the organizers had a very straightforward approach: if you are sitting at the same desk, you are a team. We were a team of three at first, later joined by a fourth, and we decided to let AI name us. We became Yabba Dabba Data. What do you know, the name actually stuck! I had the pleasure of teaming up with Brandyn Ewanek, Priya Pillarisetty, and Shivam Dhawan.

The Early Game

The challenges were a good mix of DevOps and Generative AI. We quickly realized that trying to pair-program our way through everything wouldn’t work, so we split our efforts. I picked up the first task, which was a DevOps multi-agent AI system with a CI/CD pipeline.

I’m inherently lazy when it comes to repetitive tasks, so diving into the CI/CD pipeline was right up my alley. The Lambda function was throwing an error, and the job was to figure out the root cause by inspecting CloudWatch logs and fixing it. Eventually, this meant updating the code in an S3 bucket to get the pipeline deploying correctly.

The Surprise on the Leaderboard

I was deeply engrossed in configuring S3 buckets and IAM permissions when my teammate, Brandyn, tapped my shoulder and mentioned we were first on the leaderboard.

I don’t have the first clue how we managed it so quickly, but when I looked up at the screen, there we were. We occasionally dropped to second or third place, but we always managed to claw our way back up. By the end of the session, we had built an astonishing 100-point lead over the next team.

Anatomy of the Challenges

Over the next two hours, we managed to fully complete six challenges without using a single clue—which was very refreshing! Here is a breakdown of what I worked on:

• Deep thought DevOps multi-agent AI: As mentioned, this involved fixing a broken CI/CD pipeline that was deploying to a Lambda function, relying heavily on CloudWatch for debugging.
• EBS cost attribution compliance: This was essentially a compliance check. I had to create a rule to ensure that all EBS volumes had a specific set of tags associated with them for cost tracking.
• Egress networking: A very quick 80-point task. The goal was to block IPv6 ingress into an EC2 container while allowing egress.
• Intelligent content transformation pipeline: This was fascinating. I used Amazon Bedrock to define agents and a knowledge base, fixing permissions so it could access training data in S3. This data was used to define guardrails. Ultimately, it formed an end-to-end function that converted newly uploaded reports in an S3 bucket into an audio playback.

The One That Got Away

I was so close to completing a seventh challenge: a legal document discovery assist application. This challenge provided 20 files describing a legal case, and the task was to configure an application to answer questions based on this data source using Bedrock.

One of the things that tripped me up was that the application defaulted to a region we didn’t have access to. I went down the wrong track of trying to define IAM policies to access that region, completely missing that I just needed to change the default region in the application code.

Once I fixed that, it started working. But then something weird happened—the knowledge base in Amazon Bedrock just disappeared. It was very weird, and with just two minutes left on the clock, I didn’t have time to recreate it. I still managed to get 90 out of 150 points for the parts I did complete.

Wrapping Up

With six challenges fully completed, we racked up 1,020 points. Not bad for a team of strangers who had only met a few hours prior.

This overall experience was amazing and a massive confidence boost. If you ever get the chance to attend one of these, I highly recommend it. You can check out the meetup event right here.

That’s a wrap. I’m definitely looking forward to more jam sessions in the future.