Internship Machine Learning for Cell Dynamics

SFBI; Montpellier; On-site; EUR 7,800 - 13,000; Full time

Job summary

SFBI in Montpellier invites students in mathematics, CS or ML to an internship focused on machine learning for cell dynamics in biomedicine.

The project covers variational autoencoders, probabilistic modelling, and dynamical systems to understand cell-state transitions and responses to perturbations.

Location is Montpellier, France; deadline 15 November 2026; host teams LPHI, CNRS, INSERM and IGF collaborate.

Qualifications

• Must be a student in related fields.
• Interest in biomedicine and single-cell data.
• Experience with Python and ML is beneficial.

Responsibilities

• Develop ML methods based on VAEs for cell dynamics.
• Reconstruct dynamical systems from high-dimensional gene data.
• Collaborate with host and partner teams at LPHI and IGF.

Skills

Applied mathematics

Computer science

Machine learning

Education

Student in applied mathematics, CS, or ML

Tools

Variational Autoencoders

Job description

Internship Machine Learning for Cell Dynamics

Description

Machine Learning for Cell Dynamics

In biomedicine, single-cell RNA sequencing (scRNA-seq) measures the activity of thousands of genes in individual cells,
producing high-dimensional datasets that capture cellular diversity and transitions between cell states.
In this internship we aim to reconstruct a dynamical system from noisy snapshots of thousands of genes,
when the underlying biological processes are only partially observed.
The challenge is to compress these data into a low-dimensional representation while preserving
the structure of cell-state transitions, infer the stochastic dynamics governing these transitions,
and identify a small set of variables and interactions that explain the observed behaviour.
This requires going beyond standard prediction methods to build models that are both interpretable
and capable of predicting responses to perturbations.

The internship will develop machine-learning methods based on variational autoencoders (VAEs),
combining deep learning, probabilistic modelling and dynamical systems. Applications include
cell differentiation and glioma progression, with the goal of understanding how cells switch
between states and how these transitions change during tumour progression and treatment resistance.

Details about this internship are available upon request.

Host team — LPHI, University of Montpellier, CNRS, INSERM
An interdisciplinary team specialising in multiscale dynamical modelling of biological systems
(deterministic and probabilistic approaches), machine learning and artificial intelligence.

Partner team — IGF, Montpellier
A team specialising in gliomas, combining clinical data from neurosurgery (Prof. Hugues Duffau)
with multi-omics data and molecular neuroscience research (Prof. Jean-Philippe Hugnot).

Location: LPHI, University of Montpellier, CNRS, INSERM, Montpellier, France
Profile: Students in applied mathematics, computer science, machine learning or related fields.
Deadline: 15 November 2026.