The first two AWS Summit qualifiers in London and Bengaluru have come and gone. Let me share my experience of participation, the results and how you can join.
AWS AI League
AWS AI League builds on the legacy of fun and engaging educational products and programs by Amazon Web Services. It combines numerous LLM and agentic tasks in gamified challenges which users solve as you acquire real-life applicable skills.
The best players meet at the end of the season at the AWS re:Invent in Las Vegas to compete for prizes, bragging rights and glory.
Visit the AI League Page to learn how you and your organisation can join the fun and supercharge building the skills for the future.
Join the AWS AI Community to meet curious minds and learn (and compete) with each other.
First two finalists of the AWS AI League 2026
Join me in congratulating the winners and finalists of the opening compatitions in London and Bengaluru.
London results:
- Josh Bayne
- Tomasz Ptak
- Mark Ross
Bengaluru results:
- Mayur Dhavan
- Darshan G.
- Antrixsh Gupta
The finalists received generous AWS credit packages. The winners also qualified to the AWS AI League 2026 finals during the AWS re:Invent 2026 taking place later this year in Las Vegas. The reward includes an expenses paid trip and accommodation, and tickets to the AWS re:Invent conference!
As a London finalist I will only very qietly rant that we didn’t get the belts for the second or third place. They really look awesome!
AWS Summit Bengaluru
I haven’t had the pleasure to participate in the Summit. Fortunately, Mayur Dhavan has shared his experience which I am pleased to share below:
The Personal Journey to the Summit
From the Farm to the Cloud: I am a 22-year-old MCA student and AI Automations & Dev Engineer based in Pune, originally hailing from a farming family in Ahilyanagar, Maharashtra.
To attend this AWS Summit, I boarded an early morning flight from Pune to Bengaluru on April 23rd. It was my first time ever stepping onto an airplane, making this entire journey to a global championship even more surreal.
The Challenge & My Architecture
The challenge was a grueling, gamified test of Generative AI engineering: building a completely autonomous AI agent using Amazon Bedrock to navigate a treacherous dungeon, solve complex coding/logic challenges, and maximize the score.
To pull this off, I architected a bulletproof backend system:
- AgentCore Memory: Forced the agent to recall backward-encrypted codes across long game loops.
- Bedrock Guardrails: Tuned filter strengths to actively dodge conversational traps without breaking the game API.
- Amazon SageMaker & AWS Lambdas: Utilized SageMaker Code Editor to quickly write, test, and deploy highly optimized Python pathfinding algorithms. This prioritized high-value loot while actively treating traps and un-keyed doors as impassable walls.
The Qualifiers & The Final Showdown
The energy in the room was intense with 250 incredibly talented builders. I successfully battled into the Top 3 to secure my spot, alongside Darshan Gandhi (1st) and Antrixsh Gupta (2nd).
For the live finale, the leaderboard was wiped completely clean to zero. It was a 3-round gauntlet featuring entirely different map challenges and dynamic scoring.
The hardest part of the finale was that no backend code tweaks were allowed. It all came down to the robustness of our architecture and raw, real-time situational prompting.
Relying on the foundation I built, I took the #1 spot right out of the gate in Round 1 and held the lead all the way through Round 3 under immense pressure.
Emerging as the undisputed winner, claiming the Championship Belt, and securing a fully sponsored ticket to compete at the global AI Agents challenge at AWS re:Invent in Las Vegas is a dream come true.
Thank you for sharing Mayur, and once more congratulations!
You can also enjoy the LinkedIn posts from the organisers and the champion himself.
AWS Summit London
I have the pleasure of participating in the AWS Summit London and the luck to qualify into the final. Let me share my experience with you so that you know what to expect and how to prepare.
Note: Each summit is a bit different. You may have more or fewer seats, more or less time, even the challenge will evolve with time as it usually does. What is certain is that you will be facing the same challenge and rules as everybody else in a given event.
The challenge we are working to solve in the AI League 2026 involves building an orchestrator agent that controls a dungeon maze game character. Under the constraint of time, lives and challenges, the caracter has to complete game levels and gather as many points as possible.
The competition is taking place in an AI League service linked with a workshop account. You solve the tasks using a simplified agent building a service but still have access to the account which allows you to investigate the results of your work. There are two phases: the qualifier workshop and the final.
AWS AI League workshop
Even if you don’t want to participate competitively, I strongly encourage you to join the workshop as you walk out having created an actual working agent with sub-agents, memory, lambda tools and guardrails which you get to apply it to the task.
You log into the workshop portal and receive a so-called burner account that you will work with. Then you set up your access to the AI League service and follow a step-by-step guide explaining the challenge and leading you through the steps of building an agent, sub-agents and the tools.
We had around 64 spots in the London workshop and among them were also semifinalists from the 2025 AI League, Leena Kaur and Mark Ross. The game has changed slightly compared to the 2025 finals in that there were new challenges such as having to find a key with a secret and using that to open a door.
The competition part of the workshop allows you to perform a test run of your agent on a game level that you see, and then you can submit your solution to a game level which you do not know anything about. The result you get from the stealth submission is your leaderboard result.
I followed the workshop and used Kiro to aid my preparation (it’s an AI League, use of AI is expected pretty essential to participate). In my preparation I have considered what challenges I might be experiencing in the hidden level and tried to prepare for that. I also kept testing my solution in various scenarios, trying to understand where I might still improve.
The score challenges that I have not managed to approach enough were about the output tokens (the fewer output tokens your agent uses the better the score) or the LLM fine-tuning (you get bonus points if you manage to). They both show a promise that can make a difference in your competition, I simply ran out of time that I could dedicate to it.
Quite quickly I managed to submit a decent result which put me on top of the leaderboard but have stagnated and got quickly overtaken first by Mark who effortlessly jumped to the first and never gave it away - you could clearly see the experience he’s built in his most previous AI League starts. Soon after I got overtaken by Josh, and the fourth participant, Vedant. Feeling the pressure I started looking for possible problems in my solution and found that I was a bit too restrictive in some of my optimisations, meaning that I wasn’t giving my agent a chance to do more. I corrected this and in a last dash managed to get a more comfortable advantage which lead me into the finals from the third spot.
Competitions like this require some preparation and determination but also depend on luck. Having participated in the AWS DeepRacer League I know well that getting into a top three is not just about knowing your stuff. Vedant in 4th and Enough in 5th were also solid participants. In reality however all the workshop participants are winners in that they take home practical knowledge of building agents and tools, and the curiosity to build on.
My chances of doing this well again are pretty slim, as the competition takes off we will see more strong players emerging. The more the merrier, and in the mean time let me share the experience of th final.
AWS AI League final
In the final all three participants start with a clean slate. Their qualifying agents take on three game levels, side by side, in an attempt to again maximise their score. Contestants can see the level and can provide a small steering instruction to their agent ahead of the run.
The final showed the need for a flexible strategy. Some of the levels were loaded with traps and opportunities for point gathering but you had so little time that you could get lost on the way. The winning treasure chest offers the highest reward and usually isn’t something to miss out on.
The first level has put me ahead of Josh and Mark, while in the second one Josh got slightly ahead of me. In the third level we saw the abundance of opportunities, and this is where my agent took the wrong path and started struggling with the challenges while Josh’s one, steered with a confident “You got this” went straight for the chest, minimising the risks and going for the time and lives bonuses. Mark’s agent seemed a bit sluggish in the third round but was gathering solid points and almost pushed me down to the third. Luckily I defended my spot.
The emotions of competing in front of an audience, and the challenge of keeping a cool head are like nothing else. The league is fun and addictive and I expect it to get big.
I also recommend reading Josh’s post about his experience on LinkedIn.
First two tickets to Las Vegas punched
With this great achievement Josh and Mayur qualified to the finals which are taking place during the AWS re:Invent 2026 in Las Vegas. Let’s see who will join them! Next stop: Sydney, 13-14th May.
How to join the AWS AI League
You can still qualify at the AWS Summits around the world, read the Official AWS AI League Announcement for details. Also follow the AI League in the AWS AI League Space in the AWS Builder Center for updates on how you can participate in the League and win tickets to Las Vegas.
If you represent an enterprise organisation, contact AWS with your interest and get the ball rolling. Visit AWS AI League page for more information.
In both cases, Join and follow the AWS AI Community for information, updates and to learn together.
Update: Included Mayur’s story from Bengaluru