Team members
Ondrej Kobza - Team leader
PhD student in AI at Czech Technical University in Prague, focusing on language modeling, conversational AI, and LLM safety. Part of the winning Alquist team in Alexa Prize Socialbot Grand Challenge 4 and led the team to third place in Challenge 5. Placed second as Team Lead in the 2025 Amazon Nova AI Challenge. Primary role: building the reinforcement-learning pipeline, covering state-of-the-art RL algorithms, reward functions, and managing AWS deployment
Ivan Dostal
Master's student at CTU Prague with extensive industry experience delivering end-to-end LLM solutions. Member of the 2nd-place team in the 2025 Amazon Nova AI Challenge, where he designed guardrail classifiers, synthetic data pipelines, and automated red-teaming tools. Research interests: mechanistic interpretability and AI safety. Primary role: developing the run-time security module and leading comprehensive safety evaluation.
Adam Černý
Completing his Data Science master's at CTU Prague, specializing in NLP with expertise in LLM training, prompt engineering, and evaluation. Placed 2nd in the 2025 Amazon Nova AI Challenge with team AlquistCoder. Current research explores agentic LLMs in retail; plans to pursue PhD in safe agentic LLM applications. Primary role: developing the Orchestrator system connecting architecture modules via AWS.
Jan Chleboun
First-year PhD student at CTU Prague. Won first place in a competition for improving LLM pre-training efficiency organized by BottlecapAi. Current research explores techniques for increasing synthetic data variability and improving knowledge distillation in LLMs. Responsible for: synthetic data generation pipeline, code generation, and code test generation.
Matěj Klouček
Final-year master's student in Artificial Intelligence at CTU Prague. Part of the CTU NLP research group, working as an AI researcher and engineer focused on LLM-based systems, RAG pipelines, and model evaluation. Has several years of software engineering industry experience. Focus: developing the orchestrator and implementing model-alignment techniques such as SFT and DPO.
Stanislav Khvedynich
Bachelor student in Applied Informatics at CTU Prague with extensive software engineering expertise. Works with the NLP research group at CTU, studying LLM-based systems and developing MCP agents, RAG systems, and classifiers. Contributing to: development of the Orchestrator system.
Faculty advisors
Jan Šedivý
Three decades of IT industry experience leading global R&D projects; holder of 19 US patents. Former researcher and research manager at IBM Thomas J. Watson Research Center (1992-2008) and Technical Lead Manager at Google (2008-2010). Currently leads the NLP group at CTU-CIIRC.
Sebastian Garcia (2025 challenge only)
Assistant Professor and security researcher specializing in applied machine learning for cybersecurity. Founded the Stratosphere Laboratory at Czech Technical University. Co-founded MatesLab hackspace and the Independent Fund for Women in Tech.
Additional 2025 challenge members
Maria Rigaki
PhD student in adversarial applications of machine learning and AI in cybersecurity at CTU Prague. Worked on LLM-based agents for planning and automating network security attacks. Developed and deployed a prompt injection challenge for university students.
Muris Sladić
PhD student in defensive applications of Generative AI and LLMs in cybersecurity with focus on cyberdeception at CTU Prague. Developed LLM-based honeypots simulating various systems (SSH, MySQL, POP3, HTTP) and an Attacker LLM capable of finding vulnerabilities in Linux shells.
