Missions

The successful M/F candidate will be integrated into the “Interpretable Foundation Model for Time Series” project within IRMA's MOCO team. He/she will contribute to the development of foundation models for multidimensional time series analysis, with a particular focus on prediction, classification, anomaly detection and interpretability of results. The project aims to exploit the capabilities of large language models (LLMs) to improve time series analysis in fields such as finance, biology and dynamical systems.

Activities

Development and adaptation of foundation models for time series analysis. Design and implementation of machine learning methods for prediction, classification and anomaly detection. Use of synthetic data generation techniques to enrich training datasets. Collaboration with research teams to integrate the models developed into concrete applications. Participation in scientific publications and presentation of results at international conferences. Possible supervision of students or trainees involved in the project.

Skills

xpertise in machine learning, particularly in language models (LLMs) and time series processing. Mastery of machine learning tools and libraries (PyTorch, TensorFlow, etc.). Experience in scientific programming (Python, C++, etc.) and high-performance computing (GPU, parallelization). Ability to work in a team and collaborate with researchers from different disciplines. Good scientific communication skills, both written and oral. Knowledge of synthetic data generation methods and their use in machine learning. Constraints and risks: The project does not present any major constraints or risks. The main challenges will relate to the technical complexity of the models to be developed and the integration of results into concrete applications. The candidate will need to demonstrate autonomy and scientific rigor to successfully complete the various stages of the project.

Work Context

The working time for this position is 50%. The Institut de Recherche Mathématique Avancée (IRMA) is a Joint Research Unit located on the Esplanade campus, easily accessible by public transport. Close to numerous shops and with an administrative restaurant just 5 minutes' walk away, IRMA offers a pleasant and stimulating working environment. With an established reputation in mathematics, both nationally and internationally, IRMA covers a broad scientific spectrum, from fundamentals to industrial applications. The institute has a staff of around 160, including around 100 permanent members, as well as an administrative department, a library and an IT department staffed by 4 engineers. Research at IRMA is organized into 7 teams, including one dedicated to Mathematical Modeling and Scientific Computing (MOCO team - MOdélisation et COntrôle ). IRMA is also home to an INRIA (Institut national de recherche en sciences et technologies du numérique) team and two platform structures, Cemosis (Centre de modélisation et de simulation de Strasbourg) and CeStat (Centre Statistique de Strasbourg), fostering links with the socio-economic world.