Constructing clever brokers to deal with complicated, real-world duties might be daunting. Moreover, slightly than relying solely on massive, pre-trained basis fashions, organizations typically have to fine-tune and customise smaller, extra specialised fashions to outperform them for his or her particular use circumstances. The AWS AI League offers an progressive program to assist enterprises overcome the challenges of constructing superior AI capabilities by way of thrilling competitions that drive innovation in agentic AI and mannequin customization.
In 2025, the primary AWS AI League competitors captured the eye of builders, knowledge scientists, and enterprise leaders globally. They got here collectively to resolve urgent issues utilizing the newest AI instruments and strategies. The grand finale at AWS re:Invent 2025 was an thrilling showcase of their ingenuity and abilities. Cross-functional groups from main organizations competed head-to-head, demonstrating their means to craft efficient prompts, fine-tune fashions, and construct highly effective AI brokers.
Congratulations to our 2025 AWS AI League Champions! After intense competitors amongst these three distinctive builders emerged victorious, sharing a $25,000 prize pool:
- 1st Place: Hemanth Vediyera from Cisco
- 2nd Place: Ross Williams from Aqfer
- third Place: Deepesh Khanna from Capital One
Determine 1: Left to proper: Ross, Hemanth, Deepesh
This publish explores how the AWS AI League program can be utilized to host AI competitions that may assist members expertise mannequin customization and agent constructing ideas, apply these to deal with real-world enterprise challenges, and showcase their progressive options by way of partaking, game-style codecs. We spotlight the brand new agentic AI and mannequin customization challenges, the place enterprises can apply to host inner tournaments utilizing AWS credit, and builders can compete at AWS occasions.
To get began, go to the AWS AI League product web page.
What’s the AWS AI League Championship?
The AWS AI League expertise begins with a hands-on, 2-hour workshop led by AWS consultants, adopted by self-paced experimentation. The journey culminates in a fascinating, gameshow-style grand finale, the place you showcase your AI creations and options to handle urgent enterprise challenges. The next determine exhibits these three steps.
Determine 2: AWS AI League Championship steps
Constructing on the success of the 2025 program, we’re excited to announce the launch of the AWS AI League 2026 Championship. This yr, the competitors options two new challenges that enable members to essentially put their AI abilities to the take a look at:
- The agentic AI Problem lets you construct clever brokers utilizing Amazon Bedrock AgentCore. Rivals craft custom-made agent architectures to deal with real-world enterprise issues.
- Complementing the agentic AI Problem, the mannequin customization Problem now makes use of the newest fine-tuning recipes in SageMaker Studio. Right here you customise fashions for particular use circumstances.
For the 2026 AI League championship, the prize pool doubles to $50,000, with tracks catering to builders at completely different talent ranges – from learners to superior practitioners.
Construct clever brokers with the agentic AI problem
The AWS AI League now options an thrilling agentic AI problem, the place you construct clever brokers utilizing Amazon Bedrock AgentCore to resolve complicated issues in a dynamic, game-style competitors. On this problem, brokers navigate by way of a maze-like grid atmosphere, encountering numerous challenges whereas looking for a treasure chest. These challenges map to real-world use circumstances, testing the brokers’ means to deal with inappropriate content material, execute code, use a browser, and extra.
Brokers have a time restrict to traverse the map, acquire factors, and overcome the obstacles earlier than reaching the treasure chest. The extra factors they earn, the upper they rank on the leaderboard. You’ll be able to totally customise your brokers utilizing Amazon Bedrock AgentCore primitives, which lets you extra securely scale and handle production-grade brokers. You too can choose particular fashions for supervisor and sub-agents, in addition to create customized instruments akin to Bedrock Guardrails, AgentCore Reminiscence, and AWS Lambda capabilities to assist your brokers navigate the challenges. The next determine depicts the obstacles the agent should overcome whereas touring to succeed in the treasure chest.
Determine 3: AWS AI League Agentic Problem
AWS AI League offers a full consumer interface (UI) for customers to construct their clever agent options. You need to use this no-code UI to assemble multi-agent architectures and instruments, integrating numerous elements akin to Amazon SageMaker Studio CodeEditor for interactive coding of customized Lambda capabilities and instruments. This lets you totally develop and customise your agent-based options inside the AWS AI League web site, with no need to go away the atmosphere.
The next screenshots showcase the agent constructing expertise all inside the AWS AI League web site.
Determine 4: AWS AI League agent instruments
Determine 5: AWS AI League multi agent structure
All through the competitors, customers obtain real-time agent efficiency suggestions, with a big language mannequin (LLM) evaluator offering evaluation to assist with iteration. The next picture showcases how the agent is evaluated throughout challenges.
Determine 6: AWS AI League agent problem analysis
On the grand finale, the highest finalists take the stage to showcase their brokers’ capabilities in a reside, game-show format, demonstrating the ability and flexibility of agentic AI in fixing complicated, multi-step issues. The analysis standards embody time effectivity, accuracy in fixing challenges, agent planning, and token consumption effectivity. The next snapshot exhibits the ultimate spherical of the Grand Finale at re:Invent 2025.
Determine 7: AWS AI League re:Invent 2025 Grand Finale
Customise fashions to outperform bigger fashions
AWS AI League is increasing the scope of its mannequin customization problem, permitting you to make use of the newest developments in fine-tuning strategies.
You’ll be able to entry the brand new mannequin customization expertise inside Amazon SageMaker Studio, the place you need to use highly effective new coaching recipes. The purpose is to develop extremely efficient, domain-specific fashions that may outperform the efficiency of bigger, reference fashions.
The problem begins with you honing in in your mannequin customization abilities. Utilizing the instruments and strategies you’ve gotten realized, you apply superior fine-tuning strategies to assist improve your mannequin’s efficiency. After your fashions are custom-made, the true take a look at begins. The fashions are submitted to a leaderboard for efficiency evaluation in opposition to a reference mannequin. The mannequin earns factors every time the automated decide deems your custom-made mannequin’s response to be extra correct and complete than the reference mannequin’s output. You’ll be able to showcase your superior abilities, rise to the highest of the leaderboard, and probably unlock new alternatives on your organizations.
In the course of the problem, you obtain real-time suggestions in your mannequin’s efficiency from an automatic evaluator once you undergo the leaderboard. The leaderboard evaluates submissions in opposition to a reference dataset all through the competitors, offering fast suggestions on accuracy that can assist you iterate and enhance your options. The next picture showcases how an AI critique is used to judge the custom-made mannequin.
Determine 8: AWS AI League mannequin customization analysis
On the grand finale, the highest finalists exhibit their fashions’ capabilities in a reside, game-show format, showcasing their immediate engineering talents. In the course of the gameshow, the scoring consists of professional analysis the place area consultants and a reside viewers take part in real-time voting to find out which AI options greatest remedy actual enterprise challenges. The next picture showcases the participant immediate engineering view throughout a Grand Finale.
Determine 9: AWS AI League mannequin customization Grand Finale participant view
Conclusion
On this publish, we explored the brand new AWS AI League challenges and the way they’re remodeling how organizations strategy AI growth. At AWS, we’ve realized that the quickest method to spark innovation is thru competitors. With AWS AI League, builders can now showcase their AI abilities, compete and unlock innovation.
To be taught extra about internet hosting an AWS AI League inside your group go to the AWS AI League and to dive deeper into constructing clever brokers and customizing AI fashions discover AWS AI coaching catalog on AWS Talent Builder.
In regards to the authors
Marc Karp is an ML Architect with the Amazon SageMaker Service crew. He focuses on serving to prospects design, deploy, and handle ML workloads at scale. In his spare time, he enjoys touring and exploring new locations.
Natasya Okay. Idries is the Product Advertising and marketing Supervisor for AWS AI/ML Gamified Studying Applications. She is enthusiastic about democratizing AI/ML abilities by way of partaking and hands-on academic initiatives that bridge the hole between superior know-how and sensible enterprise implementation. Her experience in constructing studying communities and driving digital innovation continues to form her strategy to creating impactful AI teaching programs. Outdoors of labor, Natasya enjoys touring, cooking Southeast Asian cuisines and exploring nature trails.
