Awarded Projects

World models enable AI systems to learn internal representations for understanding, prediction, and planning. EWM advances the Joint-Embedding Predictive Architecture (JEPA) framework to create scalable, multimodal world models trained on diverse data including internet-scale video, images, text.

The project's aim is to discover new practically-usable conventional superconductors using state-of-the-art ab initio techniques.

Two-dimensional (2D) materials can assume novel, exotic condensed matter phases highly relevant for possible breakthroughs in nano- and opto- electronics, such as ultrafast (photo)transistors and detectors.

Syngens has developed a powerful AI-based DNA Design Platform to design DNA for biomanufacturing.

This project aims to replicate and extend Anthropic’s recent breakthrough in explainability (Lindsay et al., 2025) into multimodal LLMs, specifically visual LLMs such as Qwen-VL and Gemma.

The project is a three year one aiming at studying the effect of temperature, pressure and ultrafast light absorption and their implication for the structural properties in materials displaying giant quantum-anharmonic effects.

this project advocates for data-driven reward learning: models that directly infer the quality of behaviour from raw observations across a diverse range of scenarios, including both successful and suboptimal trajectories.

This project introduces MatMulFreePLM, a protein language model that overcomes the performance plateau of existing models like ESM by expanding and diversifying training datasets while optimizing computational efficiency.

Particle accelerators and light sources are main tools for discovering the fundamental constituents of our universe, the inner workings of biological organisms, and the response of matter inside stars and planets.