Benjamin Crouzier

Benjamin Crouzier

Founder & Research Engineer

Benjamin has a masters in Computer Science likes to build stuff.

Research interests: test-time computation, UED/MCC/auto-curriculum, framing problem, tokenisation, probabilistic programming, search/self-play, music-generation, generalization, scaling laws

Mohamed Osman

Mohamed Osman

AI Researcher

Mohamed has four years of work experience in developing deep learning models and multiple successful real-world deployments under his belt.

Research interests: Meta learning, reasoning, test-time learning, active inference, few-shot learning, data scarcity

Akira Yoshiyama

Akira Yoshiyama

AI Researcher

Akira was previously focused on LLM fusion at the University of Waterloo. He has a diverse background in building innovative technology solutions, from generative machine learning applications in aerospace engineering to advanced CCTV footage analysis tools for law enforcement.

Research interests: LLM fusion, generative machine learning, AI applications in security and law enforcement, computational design and optimization, autonomous systems

Dries Smit

Dries Smit

AI Researcher

Dries is a researcher specialising in reinforcement learning, multi-agent systems, and large language models, with a strong track record of applying these techniques in medical and financial domains. Dries led the team behind Laila, a fine-tuned variant of Llama 3.1 designed to assist biologists by interfacing directly with laboratory equipment. His research currently focuses on large-scale training and reinforcement learning; follow his latest work on driessmit.github.io.

Research interests: Reinforcement learning, multi-agent systems, large language models, large-scale AI training, biomedical applications, foundational model enhancement, autonomous systems, curiosity-driven learning

Dominique Garmier

Dominique Garmier

AI Researcher

Dominique has a bachelor’s in mathematics and is currently pursuing his master’s with a focus on analysis.

Research interests: reasoning, formal proofs, statistical learning theory, functional analysis, ergodic theory, dynamical systems

Isaiah Pressman

Isaiah Pressman

AI Researcher

Isaiah is a research engineer with a focus on self-play reinforcement learning and the techniques needed to train large-scale deep learning models. His background includes machine learning engineering for dynamic pricing, applying computer vision to histopathology research, and multiple 1st- and 2nd-place finishes in Kaggle competitions.

Research interests: self-play, reinforcement learning, in-context learning, reasoning models

Jerome Sieber

Jerome Sieber

AI Researcher

Jerome is an AI researcher focusing on deep sequence model architectures and reasoning models. He holds a PhD in control theory from ETH Zurich, where he worked on optimal control for dynamical systems and sequence model architectures. His research spans various domains, including robotics, dynamical systems, and artificial intelligence. Jerome has published several papers on control theory and machine learning, contributing to advancements in predictive control and fundamental understanding of sequence models.

You can find Jerome’s latest work on Google Scholar.

Research interests: sequence model architectures, reasoning fundamentals, test-time RL, model normalization, optimization techniques, dynamical systems

Stefano Viel

Stefano Viel

AI Researcher

Stefano has a master’s in data science from EPFL. There, he primarily conducted research on reinforcement learning, large language models, and computational social science.

Research interests: reinforcement learning, unsupervised environment design, meta learning, synthetic data generation, curiosity-driven learning, and large language models

Jeroen Cottaar

Jeroen Cottaar

AI Researcher

Jeroen has a PhD in Applied Physics from Eindhoven University of Technology and over 10 years of industry experience at ASML, where he architected Bayesian machine learning approaches for semiconductor metrology. He is a top-5 Kaggle competitor with multiple 1st and 2nd place finishes across diverse domains, with solutions ranging from principled Bayesian inference to genetic algorithms with custom CUDA kernels. He now aims to merge Bayesian methods with LLM reasoning approaches.

Research interests: reasoning, abstraction, test-time computation, program synthesis, Bayesian inference, Gaussian processes, scientific machine learning

Tommy He

Tommy He

AI Researcher

Tommy has a Bachelor’s in math & computer science, worked on research in cryptography and ML, and previously co-founded YC company Clarum building AI agents for investing. His previous experience spans working on nanosecond HFT systems to privacy preserving ad-tech solutions.

Research interests: reasoning, RL, self-play, interpretability, geometry of learning, mathematical foundations of learning systems

Michal Tesnar

Michal Tesnar

AI Researcher

Michal is pursuing a Master’s in Data Science at ETH, focusing on the mathematical foundations of the field and its applications to deep learning. He has conducted research on continual learning in robotics and reasoning with diffusion models. He maintains a broad focus, exploring human cognition and the impact of technology on the society.

Research interests: reasoning, large-language model post-training, long-context agentic tasks, agentic orchestration, test-time scaling, and uncertainty estimation

Victor Mercklé

Victor Mercklé

AI Researcher

Victor is a machine learning researcher with a background in reinforcement learning and deep learning. He studied computer science at ENS Lyon. His work looks at the theory of neural network training, including how the geometry of data shapes optimization landscapes in ReLU networks. He has also worked on LLM systems, agents, and search methods for games and logic problems. Check out his work on his website.

Research interests: unsupervised training, game AI, search, coding agents, mechanical sympathy, automating research