Neural Architecture Engineering
Designing next-generation neural network architectures that improve efficiency, interpretability, and scalability for deployment in mission-critical environments.
Leipzig Advanced Research Facility
Pioneering the Next Generation of Intelligent Systems
We develop advanced artificial intelligence systems that push the boundaries of machine cognition — with safety and transparency at the core of everything we do.
2019
Founded
200+
Researchers
14
Countries
83
Publications
About
Building the future of AI — responsibly
Founded in 2019, the Leipzig Advanced Research Facility operates at the intersection of computational neuroscience, machine learning, and autonomous systems engineering. Our interdisciplinary team of over 200 researchers and engineers from 14 countries works to develop AI systems that are not only more capable, but fundamentally more aligned with human values and societal needs.
Headquartered in Leipzig, Germany, LARF has established itself as one of Europe's foremost independent AI research institutions. Our work spans foundational research, applied systems development, and AI safety — driven by the conviction that the most powerful technologies demand the most rigorous oversight.
12
Active research projects across four divisions
3
Joint ventures with European research councils
47M€
Annual research budget (FY 2025)
“The most powerful technologies demand the most rigorous oversight.”
Research Divisions
Four pillars of our work
Cognitive Systems & Reasoning
Building AI systems that exhibit robust reasoning, planning, and adaptive behavior across complex, multi-domain problem spaces.
AI Safety & Alignment
Developing theoretical frameworks and practical tools to ensure advanced AI systems remain aligned with specified objectives and human values.
Autonomous Decision Systems
Creating reliable decision-making frameworks for AI agents operating in high-stakes, real-world environments with incomplete information.
Flagship Project
Project Obsidian
Launched in January 2024, Project Obsidian represents LARF's most ambitious research initiative to date. The project aims to develop a unified cognitive architecture — a single system capable of general reasoning, continuous learning, and autonomous adaptation across unrestricted domains.
Obsidian integrates breakthroughs from all four of our research divisions into a coherent, scalable framework. The system architecture is built around a novel self-modifying inference engine, enabling real-time optimization of its own reasoning processes.
The project is currently in Phase III: Advanced Integration & Stress Testing.
