Probabilistic inference based on machine learning and generative AI, applied to evolutionary genomics M/F

Organisation/Company CNRS Department Direction des ressources humaines Research Field Engineering Computer science Mathematics Researcher Profile First Stage Researcher (R1) Application Deadline 2 Sep 2026 - 17:00 (UTC) Country France Type of Contract Other Job Status Full-time Hours Per Week 35 Offer Starting Date 4 Aug 2026 Is the job funded through the EU Research Framework Programme? Not funded by a EU programme Is the Job related to staff position within a Research Infrastructure? No

Offer Description

Evolutionary genomics stands out as a prime example where scientific probabilistic modelling naturally intersects with modern generative learning approaches, opening up a particularly rich field of collaboration between life sciences, probability theory and computer science. The LISN (UMR 9015) and LBBE (UMR 5558) laboratories are international leaders in bioinformatics, evolutionary genomics and statistical modelling, equipped with state-of-the-art technological platforms (high-performance computing, genomic databases, AI tools).The CPJ will strengthen synergies between computer scientists, biologists and modellers, drawing on national and international programmes (e.g. CNRS Bioinformatics networks, collaborations with INRIA, INSERM, or European universities). It will help position France as a leader in AI for genomics, in relation to public health, biodiversity and methodological innovation.

This CPJ focuses on two complementary laboratories, LISN and LBBE, which specialise in evolutionary genomics. Their research, at the interface between machine learning and genomics, combines methodological innovations (generative models, causal inference, multi-omic representation) with biological applications (population genomics, phylogenetics, phenotype prediction). Their interdisciplinary approach brings together biologists, physicists and computer scientists to decipher evolutionary mechanisms.The CPJ will benefit from their expertise in modelling adaptive and evolutionary dynamics, as well as in big data analysis (semi-supervised learning, inference via simulation). The integration of heterogeneous data (multi-omic, environmental) will enable the development of robust and interpretable predictive models, addressing current challenges in evolutionary biology.

Scientific machine learning is undergoing a revolution driven by generative AI, which is capable of learning, often from unsupervised data, universal probabilistic models of the statistical structure of the domains under study. This transforms tasks such as prediction or classification into specific cases of inference, amplified by the power of these models. The challenge lies in extracting interpretable latent variables that reflect the underlying biological mechanisms.Evolutionary genomics, with its probabilistic models of biological processes (mutations, genome structure, biophysical constraints, selection) and its massive but noisy datasets, offers an ideal testing ground for these advances. Recent progress in AI in biology (AlphaFold, ESM, DNABert) now makes it possible to integrate sequence, structure, function and evolutionary dynamics. Thus, this field naturally draws on emerging ML approaches (simulation-based inference, diffusion models, neural ODEs/SDEs, variational autoencoders) to develop scalable, interpretable inference methods that are accessible to the scientific community.

Teaching project will be discussed depending on the university that will welcome the CPJ.

The CNRS is developing a strong policy in favor of open science. Open science consists of making research results "as accessible as possible and closed as necessary". As such, the CNRS aims to make 100% of the texts of publications resulting from the work of its laboratories accessible , in particular through deposit in HAL. The data produced must also be made available and reusable, except for specific restrictions. In addition, the guiding principles of individual evaluation have been revised in accordance with the DORA declaration, to be more qualitative and to take into account all facets of the researcher's profession.

The dissemination of the results will be done through world-class scientific productions: publications, patents, software... In addition, the results will be communicated to various targets such as scientific communities, media, decision makers, general public, schools, etc., with an adapted calendar. Specific tools may be developed such as websites, newsletters, meetings, international symposia, summer schools and conferences.

The relationship between science and society is now recognized as a full dimension of scientific activity. The project will develop this dimension in synergy with all the partners. The resulting research work will contribute to informing public decision-making. Participatory science initiatives may be initiated with actors from the project's socio-economic and cultural eco-system.

holders of a doctorate or a PhD or equivalent degree or applicants who have gained scientific. There is no restriction on the age or nationality of applicants. All CNRS positions are accessible to people with disabilities, with special arrangements for tests made necessary by the nature of the disability.