• October 2022: I am looking for outstanding and highly motivated Post-Doctorate to work on machine learning and chemical reaction predictio (molecule representation, deep learning, transfer learning). More details on the job offer are here
• September 2022: We are organizing ECML-PKDD 2022 to be held in a hybrid mode in Grenoble (France). Check the Invited Speakers, the Workshop & tutorials and the accepted papers here https://2022.ecmlpkdd.org/!
• July 2022: New paper accepted to ICML “Bregman Neural Networksn” with J. Frecon, S. Salzo and M. Pontil, toolbox
• May 2022: New paper “Physically-admissible polarimetric data augmentation for road-scene analysis” with my Alumni C. Ruffino accepted to CVIU
• December 2021: Elected as Executive Board of SSFAM, the French Society in Machine Learrning

1-year Post-doc position

Expected Starting date: as soon as possible but no later than end of January, 2023

Title: Machine Learning for chemical reaction prediction

Keywords: Machine Learning, Molecule Representation Learning, Deep Learning, Transfer Learning, chemical molecules synthesis, cyclopropanes.

Context

This position is proposed in the context of the french project CYCLIA for tailoring statistical models to optimize the asymmetric synthesis cyclopropane compounds: chemical molecules representation, prediction algorithms, evaluation framework. The project is funded by the Normandy Region. It gathers 2 research teams: the Machine learning team of LITIS lab and the Biomolecules Synthethis Team of COBRA at INSA Rouen Normandy (France).

The Post-doc candidate will join the Machine learning team of the LITIS in Rouen (France), and will have the opportunity to closely collaborate with the reknown experts in Organic and Bio-organic molecules synthesis.of COBRA. The Machine learning team is composed of 15 permanent researchers and several PhD, post-doctorates. The candidate will be part of the machine learning project of the lab whose topics include statistical learning, deep learning, representation learning, domain adaptation, graph learning…

Scientific objectives

The objective of this project is to propose new machine learning methods to predict/optimize the asymmetric synthesis cyclopropane compounds that are present in the structure of many drugs, based on the chemical reactants, the catalysts, and the experimental conditions.

Research topics of interest include but are not limited to:

• chemical molecules representation learning approaches,
• deep learning approaches,
• transfer learning,
• development of new algorithmic approaches,
• proper evaluation of designed algorithms.

Required skills

We are searching for a highly motivated candidate with:

• A PhD degree in machine learning or related domains
• Strong programming skills in languages such as Python.

Working environment

The post-doc position is located in Rouen a mid-size city of Normandy (France). The hosting research group has established expertise in relevant domains including statistical machine learning, deep learning, optimal transport, domain adaptation. The successful candidate will have the opportunity to work in synergy with a post-doctorant and a Ph.D. student involved in the CYCLIA project. The expected salary is 2172 euros NET per month, including healthcare.

Application instructions

The candidates should send:

• a CV (with the publications record),
• a motivation letter explaining the relevance of the candidate for the position
• names and contact details of at least two references.

Contacts

Gilles Gasso, gilles.gasso[AT]insa-rouen.fr
Phone: (+33) 2 32 95 99 88
Samia Ainouz, samia.ainouz[AT]insa-rouen.fr
Benoit Gauzere, benoit.gauzere[AT]insa-rouen.fr

Post-Doc

• Jordan Frecon (2019-2022), post-doc within the RAIMO project by S. Canu
• Mokhtar Alaya (2018-2020)

PhD Students

• Guillaume Mahey (2021-2024?) co-supervised by Laetitia Chapel
• Marwa Kechaou (2018-2022?) co-supervised by Romain Hérault
• Anna Stepien (2018-2022?) co-supervised by Romain Hérault
• Michel Olvera (2019-2022?) co-supervised by E. Vincent

Recent Alumni

• Cyprien Ruffino co-supervised by Romain Hérault (now at Kinetix)
• Benjamin Deguerre co-supervised by Clément Chatelain (now at Actemium PT)
• Imad Rida co-supervised by Romain Hérault (now at UTC)
• Xilan Tian co-supervised by S. Canu (now R&D Engineer)
• Gregory Mallet co-supervised by S. Canu (now R&D Engineer)
• Karina Zapien co-supervised by S. Canu (now R&D Engineer at Leboncoin)

Matrix computation

This course (in French) is mostly about numerical linear algebra.

• Introduction to matrix calculus
• Notion of numerical complexity
• Matrix Factorizations (LU, Cholesky, QR factorization)
• Solving linear system
  • Review of Least squares problem
  • Application of matrix factorizations
  • Iterative methods (Jacobi, Gauss-Seidel, relaxation)
• Algorithms for Eigenvalues and Singular values computation

Details can be found here ITI - Méthodes Numériques pour l’Ingénieur

Machine Learning

This course introduces the conventional methods in unsupervised and unsupervised learning. It provides some (convex) optimization notions. Numerical materials are in Python.

Details can be found here ITI - Introduction au Machine Learning

Data Science

The course introduces to tools for data science. The target audience is Master students aiming pluridisciplinary skills. Numerical materials are in Python.

Details can be found here MS ESD - Ingénierie des données