Senior Research Scientist - Generative models fine-tuning & post-training

S27a; Paris; Hybride; EUR 90 000 - 130 000; Plein temps

Avantages offerts par ce poste

Equity options

25 days paid time off

Résumé du poste

Karavela in Paris is seeking a Senior/Staff Research Scientist to lead the neuro-AI effort within the model development team. You will shape the scientific agenda, exploring how brain recordings can steer or fine-tune pre-trained models to align with human brain processing of language and sensory information.

You will implement training pipelines for fine-tuning on brain data, collaborate with data acquisition on stimuli, and mentor a growing team of researchers and engineers in a foundational

Qualifications

• The candidate holds a PhD (or equivalent) in ML, CS, computational neuroscience, or related field.
• Strong track record in deep learning research with publications or equivalent industry experience.
• Excellent Python and PyTorch programming/engineering skills.
• Hands-on experience fine-tuning or aligning large pre-trained models (LLMs).
• Experience with multimodal models and cross-modal representation alignment.
• Ability to define and drive an independent research agenda and mentor juniors.
• Strong communication across MRI physics, ML, and clinical applications.
• Familiarity with neural/physiological data (fMRI/EEG/MEG) is a plus.

Responsabilités

• Lead our neuro-AI effort within the model development team.
• Define and drive the scientific agenda from exploration to pipelines for fine-tuning models on brain data.
• Collaborate with data acquisition to shape stimuli and protocols supporting research.
• Build and mentor a team of researchers and engineers as the effort grows.

Connaissances

PhD/Equiv

Deep learning research

Python

LLM fine-tuning

Multimodal models

Mentoring

Cross-disciplinary communication

Neural data familiarity

Formation

PhD or equivalent

Outils

PyTorch

Description du poste

Context

Karavela is a neuroscience research company established in 2025 and currently based in Paris, France. We build a foundation model of the human brain relying on next-generation magnetic resonance imaging. To do so, Karavela is collecting a very large MRI dataset from healthy volunteers listening to naturalistic stimuli such as audiobooks, podcasts and radio programmes. This dataset is then used to train deep learning models that can reconstruct the stimuli from recorded brain activity. The goal is to leverage the representations learnt by these models in downstream applications, in particular by providing biomarkers for neurological and psychiatric diseases in clinical trials.

Karavela is currently organised into 3 teams focusing on data acquisition, model development and clinical applications.

Position summary

Karavela is looking for a Senior/Staff Research Scientist to lead our neuro-AI effort within the model development team. Your mission will be to explore how brain recordings can be used to improve generative models, in particular by leveraging pre-trained large language and multimodal models. You will investigate how neural signals recorded during naturalistic stimulation can be used to steer, adapt, or fine-tune these models, so that their internal representations better align with the way the human brain actually processes language and other sensory information. This is a foundational research role: you will define and drive the scientific agenda for this line of work, from early-stage exploration to building the training pipelines needed to fine-tune and post-train pre-trained models on brain data at scale. You will work closely with the model development team, as well as with the data acquisition team to help shape the stimuli and protocols that best support this research direction. As the effort grows, you will help build and mentor a team of researchers and engineers around it.

Qualifications

The ideal candidate will have:

• a PhD (or equivalent research experience) in machine learning, computer science, computational neuroscience or a related field;

• a strong track record of research on deep learning, with publications at top-tier venues (e.g. NeurIPS, ICML, ICLR) or equivalent industry experience;

• excellent programming and deep learning engineering skills, particularly in Python and PyTorch;

• hands-on experience fine-tuning, post-training or aligning large pre-trained models (e.g. LLMs), including techniques such as instruction tuning, RLHF/DPO, or parameter-efficient fine-tuning;

• experience working with multimodal models (e.g. combining text, audio, or vision modalities), and an understanding of how to align or fuse representations across modalities;

• ability to define and drive an independent research agenda, and to mentor junior researchers and engineers;

• strong communication skills and comfort working across disciplines, from MRI physics to machine learning to clinical applications;

• familiarity with neural or physiological data (fMRI, EEG, MEG) and/or computational neuroscience is a strong plus.

Benefits

This position offers:

• open-ended contract (CDI);

• competitive compensation, depending on experience;

• equity options;

• 25 days of paid time off per year, in addition to French reduction of working time (RTT);

• full-time position in Paris, partial remote work is possible.