Our Team

Benjamin has a masters in Computer Science and applied ML to quant finance previously.

Research interests: test-time computation, active inference, system2 reasoning, framing problem, tokenisation, probablistic programming, search/self-play, music-generation

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

Toby has a strong background in applied mathematics and extensive experience in leading innovative technology projects. He previously co-founded and served as CTO of Inspire Robotics, an edutech company delivering STEM education to thousands of students annually. He has also led engineering projects in satellite systems and humanoid robotics, significantly enhancing performance through the integration of AI.

You can find Toby's latest work with Tufalabs here: https://arxiv.org/abs/2503.00735
(HN thread, Twitter thread, Reddit thread)

Research Interests: optimal model routing, scaling inference-time compute, mixture of expert agent models, reinforcement learning with large language models (LLMs), self-improving architectures, domain-specialized models, autonomous robotics systems.

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.

Kevin is an AI researcher specializing in deep learning, computational neuroscience, and software engineering. He holds a master's from ETH Zurich, where he developed advanced machine learning frameworks and contributed significantly to research in recurrent neural network architectures. Kevin's work spans from developing open-source software for computational neuroscience to engineering robust algorithms for natural language processing and distributed computing.

Research Interests: deep learning architectures, sparse recurrent neural networks, computational neuroscience, natural language processing, distributed training methodologies, synthetic data generation for language models.

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. He has extensive experience in designing advanced multi-agent frameworks, improving foundational models, and leading impactful projects. Notably, 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.

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 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