Introducing Karavela AI - Data and models to understand the brain

We are very excited to announce the creation of Karavela, a research company based in Paris with the goal of acquiring a very large dataset of brain activity.

Our mission

The brain remains one of the most complex systems in medicine and one of the most underserved. Neurological and psychiatric disorders represent some of the biggest unmet medical needs globally, yet CNS drug development suffers from one of the highest failure rates in the industry. The main challenges include:

• a lack of robust functional biomarkers ;
• subjective and slow clinical endpoints ;
• high patient heterogeneity ;
• limited tools to stratify patients or monitor response.

At Karavela, we believe the future of neuroscience requires a step-change in how we model the brain. Focusing on AI-native approaches, we are firm believers that richer and larger data are a necessary step towards more precise models of the brain's activity.

Our solution

We are training a foundation model of the brain, based on hundreds of hours of task-based fMRI collected under rich perceptual and cognitive protocols. Once trained, this model can be used to develop numerous solutions:

• detect response or resistance to treatment early ;
• stratify patients based on brain function, not just symptoms ;
• generate quantitative, functional biomarkers for trials and care ;
• create the basis for next-gen brain-computer interfaces.

Unlike standard approaches that rely on resting-state brain activity analysis, Karavela’s platform captures active brain function elicited by well-characterized yet rich stimuli. It uniquely leverages both image data and raw MRI signal (k-space) to provide finer brain models.

Our co-founding team

Karavela takes advantage of the respective domains of expertise of its 3 co-founders:

• Bertrand Thirion, PhD, is the head of science at Inria Saclay. He co-founded scikit-learn, led the acquisition of the Individual Brain Charting dataset, a collection of 1000+ hours of functional MRI data. With 120k+ citations and 60+ h-index, he is a highly-regarded researcher in the fields of machine learning and computational neuroscience.
• Charles Maussion specialised in ML applied to clinical research. His research involves leveraging medical imagery to train models providing biomarkers for oncology-related diseases. He was the director of biomarkers R&D at Owkin, and led the team’s technical developments while interfacing with pharma clients.
• Alexis Thual, PhD, a doctorate in ML applied to neuroimagery. His research focuses on training models decoding perception, and transferring them across participants. He worked in Meta’s Brain & AI team, and previously co-founded Arkhn, a startup specialising in healthcare data integration.

Together, the goal is to create a uniquely large dataset optimised for decoding purposes, in order to best identify patterns in brain activity that could in turn be useful in clinical trials.

Collaborate

We are a passionate team of neuro-AI experts. Whether you want to join us, collaborate with us on research projects or support our journey, feel free to reach out to us!