CATALYST Lab

Our Mission

The Kachman CATALYST Lab operates at the cutting edge of interdisciplinary science, coalescing ideas and approaches from machine learning, complexity theory, network science, and deep numerical modeling. We tackle cutting-edge problems in artificial intelligence and the natural sciences, bridging the gap between theoretical understanding and practical applications.

Our mission is to advance the frontiers of AI by developing novel approaches that combine insights from game theory, complex systems, chemical sciences, and cognitive science. We believe that the most impactful breakthroughs emerge at the intersection of diverse disciplines.

Research Philosophy

At CATALYST Lab, we embrace a collaborative, interdisciplinary approach to research. Our philosophy is built on several core principles:

Interdisciplinary Innovation: We believe the most exciting discoveries happen when different fields collide. Our team brings together expertise from AI, chemistry, game theory, complex systems, and cognitive science.
Rigorous Theory Meets Practical Impact: We develop mathematically grounded methods that solve real-world problems, from understanding LLM reasoning to accelerating chemical discovery.
Open Science & Reproducibility: We are committed to open-source code, transparent methodology, and reproducible research. Check out our resources page for publicly available code and datasets.
Student-Centered Mentorship: We maintain an open-door policy and prioritize deep, hands-on mentorship. Our students are encouraged to pursue ambitious ideas and collaborate across projects.

Lab Culture

CATALYST Lab fosters a welcoming, collaborative environment where creativity and intellectual curiosity thrive. We value:

Collaboration over Competition: Team members regularly collaborate across projects, share insights, and support each other's research goals.
Diversity of Thought: We actively seek perspectives from different backgrounds, disciplines, and career stages.
Work-Life Balance: We recognize that great research requires sustainable work practices and time for reflection.
Intellectual Freedom: While we have core research themes, students are encouraged to explore unexpected directions and follow their curiosity.

Research Impact

Our work has contributed to multiple domains:

AI Safety & Reasoning: Developing benchmarks and methods for understanding strategic reasoning in large language models.
Autonomous Chemical Discovery: Creating AI systems that accelerate the discovery of new molecules and materials.
Game-Theoretic AI: Building efficient methods for value attribution and equilibrium computation in multi-agent systems.
Neural Network Training Dynamics: Uncovering fundamental principles that govern how deep learning systems learn and generalize.

Our research appears in top-tier venues (NeurIPS, ICML, AAMAS, Physical Review Letters, JCIM) and has been adopted by industry partners and research labs worldwide.

Collaboration Opportunities

We actively seek collaborations with academic labs, industry partners, and interdisciplinary researchers. If you're interested in working together on:

LLM reasoning and multi-agent systems
AI for chemical discovery and materials science
Game theory applications in AI
Training dynamics and optimization theory

Please reach out via our contact page. We're always excited to explore new partnerships and research directions.

Location

We are located in the beautiful city of Nijmegen at Radboud University, distributed between the Artificial Intelligence Department, Data Science program, and Institute for Molecules and Materials.