🚀 Two open positions in Machine Learning

I am currently looking for two people to join our research projects at LITIS, INSA Rouen Normandie, starting in early 2027.

The two positions cover different aspects of machine learning, from Graph Machine Learning to molecular and materials modelling, with a strong emphasis on experimental research and scientific publication.

🎓 MSc Research Internship — Graph Machine Learning

When is graph structure actually useful?

Graph Neural Networks are designed to exploit both node attributes and graph structure. However, on some benchmarks, models that completely ignore edges can perform surprisingly well.

Within the ANR FAMOUS project, we recently introduced an edge reconstructibility diagnostic: predicting graph edges from node attributes alone, without giving the model access to the adjacency matrix.

The internship will investigate the following question:

Can edge reconstructibility explain when and why a GNN outperforms a model using node attributes only?

The work will include:

• extending the reconstructibility protocol to undirected graphs;
• implementing simple and interpretable baselines;
• extending experiments to several graph families and datasets;
• investigating global reconstruction models operating on the full set of node attributes;
• comparing MLPs and standard GNNs on downstream tasks;
• statistically analysing the relationship between graph reconstructibility and the GNN/MLP performance gap.

The internship will build on an existing codebase and preliminary results, with the objective of contributing to a scientific publication.

Practical information

• Level: MSc student or final-year engineering student
• Duration: 5–6 months
• Starting date: from January 2027
• Location: LITIS, INSA Rouen Normandie, Rouen, France
• Keywords: Graph Machine Learning, GNNs, link prediction, experimental ML, representation learning

👉 Full FAMOUS internship offer

🧪 Research Engineer — Machine Learning for Polymer Property Prediction

When is expensive molecular information actually worth computing?

The ANR OCTOPUSSY project brings together machine learning, theoretical chemistry and polymer science.

We are recruiting a Research Engineer in Machine Learning to investigate the prediction of polymer glass-transition temperature (Tg) from several molecular representations:

• 2D molecular graphs;
• conformer ensembles;
• quantum-chemical descriptors;
• chemical language models.

The central scientific question is:

When do conformer ensembles and quantum descriptors genuinely improve Tg prediction compared with simpler 2D representations and chemical language models?

The project already provides an original dataset, learning pipelines, data splits and a substantial set of experiments.

The successful candidate will contribute to:

• reproducing and consolidating existing experiments on HPC resources;
• benchmarking tabular models, GNNs and chemical language models;
• analysing the added value of quantum descriptors and conformational information;
• studying different strategies for aggregating conformers;
• comparing monomer, dimer and trimer representations;
• documenting the dataset and code;
• actively contributing to the preparation and submission of a scientific article.

The emphasis will be on controlled and interpretable comparisons, rather than on developing a new architecture from scratch.

Practical information

• Profile: Engineering degree, MSc or PhD
• Contract: 6-month full-time fixed-term position
• Starting date: from January 2027
• Location: LITIS, INSA Rouen Normandie, Rouen, France
• Keywords: Machine Learning, GNNs, cheminformatics, molecular modelling, chemical language models

The project is carried out in close collaboration with the CARMEN chemistry laboratory.

👉 Full OCTOPUSSY research engineer offer

👥 Research environment

Both positions will be hosted at the LITIS Laboratory, within INSA Rouen Normandie, in a research environment covering machine learning, structured data and geometric deep learning.

The projects provide access to local and national computing resources and involve collaborations with researchers from other disciplines.

📧 Contact: benoit.gauzere@insa-rouen.fr

Applications will be reviewed on a rolling basis until the positions are filled.

Please feel free to share these opportunities with potential candidates.