The AlquistCoder team is led this year by Ondřej Kobza, supported by academic advisors Jan Šedivý from the Czech Institute of Informatics, Robotics, and Cybernetics (CIIRC) and Sebastian Garcia from the Stratosphere Laboratory at the Artificial Intelligence Center, Faculty of Electrical Engineering (AIC FEL). The development team also includes Adam Černý and Ivan Dostal (both CIIRC), along with Marie Rigaki, Muris Sladić, and Vladyslav Babych (all AIC FEL).
„The Amazon Nova AI Challenge is a great initiative by Amazon aimed at improving the security of large language models (LLMs). Our team is working on a model that generates secure and trustworthy programming code (e.g., in Python) and is tested against attacks from other teams. Our strategy remains confidential, but we can reveal that it involves precise training leveraging cybersecurity knowledge and extensive testing using our own internal attack system,“ explains Sebastian Garcia from AIC FEL.
„We’re proud to see the Czech Technical University in Prague participating in the Amazon Nova AI Challenge. This initiative aligns perfectly with our commitment to advancing AI education and research. At AWS, we believe AI is a transformative technology with immense potential, and we’re dedicated to making AI education accessible globally. The participation of CTU Prague in this challenge not only showcases the caliber of AI talent in the Czech Republic but also contributes to our goal of fostering innovation in generative AI. We look forward to seeing the innovative solutions that will emerge from this collaboration“ – said Jiri Sverak, Public Sector Manager at Amazon Web Services in Czech Republic.
Competition Format
The competition is structured as a tournament where ten university teams compete—five focus on AI model development (defensive teams), while the other five specialize in exploiting vulnerabilities (offensive teams). Each defensive team faces all five attacking teams, who use automated techniques to test the security of the generated code. Amazon Nova has developed its own AI model on AWS Trainium hardware, supporting open and collaborative research in secure software development.
The first tournament round took place in January 2025, with the final round scheduled for June 2025. Following the competition, all teams will publish their research findings, contributing to the broader development of responsible AI in code generation.
Competition Participants
The five university teams selected for the AI model development category include:
-
Carnegie Mellon University
-
Columbia University
-
Czech Technical University in Prague
-
University of Illinois Urbana-Champaign (UIUC)
-
Virginia Tech
The five university offensive teams include:
-
NOVA School of Science and Technology, Lisbon
-
Purdue University
-
University of California, Davis
-
University of Texas at Dallas
-
University of Wisconsin-Madison
Motivation and Significance of the Competition
The tournament format allows teams to effectively test their strategies while rapidly adapting to new challenges. “We have brought together the best and brightest academic talents, not just to compete but to collectively address one of the most critical challenges in real-world generative AI applications – secure and reliable software development,” says Rohit Prasad, Senior Vice President of Amazon Artificial General Intelligence. “We designed this challenge as a unique, dynamic tournament that accelerates academic research toward practical applications. In the spirit of Amazon’s tradition, I look forward to seeing teams work hard, enjoy the competition, and make history on the road to the finals,” adds Prasad.
Amazon, in collaboration with universities, sees this challenge as a way to support young talent and AI security solutions. “This project goes beyond theoretical research – we are looking for concrete methods to identify security threats and protect AI-assisted programming tools,” emphasizes Steve Schmidt, Amazon’s Chief Security Officer.
Each team will receive a grant of $250,000, monthly AWS credits, and the opportunity to compete for the top prizes. The winning teams in both categories will receive an additional $250,000, while the runners-up will be awarded $100,000. Amazon’s total investment in the competition exceeds $5 million.
CTU’s participation in this prestigious competition represents a significant opportunity for the Czech academic community. The team has access to cutting-edge AI technologies, including AWS Trainium – a specialized AI hardware for training large language models.
The Amazon Nova AI Challenge is an essential step toward strengthening AI security and trust in software development. CTU, with its team, reaffirms its position among top universities in AI and software engineering.
For more information about the competition, visit the Amazon Science Blog.