FTP - Postdoctoral Researcher in World Models, JEPA & AI Forecasting (M/F/X)
Software Engineering, Data Science
Context & Environment
-
You join the Power Technical Line within OneTech, as part of the Forecast team within IPM (Integrated Power Management). The team develops digital solutions supporting forecasting activities for the Power business segment, particularly in the fields of meteorology, renewable energy generation, and climate data analytics.
As part of this postdoctoral research project, you contribute to the exploration of so-called World Model architectures, with a particular focus on the JEPA (Joint-Embedding Predictive Architecture) approach, developed in the field of self-supervised learning. The objective is to assess these models' ability to learn robust representations of complex phenomena and improve predictive performance across a range of industrial use cases.
The primary application domain involves forecasting using time series of satellite imagery and meteorological data. The research may also be extended to other topics such as anomaly detection, predictive maintenance, climate data analysis, and the exploration of foundation models applied to OneTech's business activities.
Activities
-
Your main responsibilities include:
- Conducting a literature review of World Models, JEPA, and self-supervised learning approaches;
- Developing and evaluating predictive architectures based on satellite image time series;
- Designing models capable of learning robust representations in the presence of uncertainty;
- Evaluating the performance of the JEPA approach for weather and energy forecasting applications;
- Applying these approaches to complementary use cases related to anomaly detection and predictive maintenance;
- Exploring the contribution of these architectures on climate datasets and other industrial datasets;
- Contributing to the development of foundation models tailored to OneTech's needs;
- Presenting research findings to technical teams and scientific communities;
- Contributing to scientific publications, conference presentations, and the dissemination of methodologies within OneTech.
Candidate Profile
-
You hold a PhD awarded within the last three years in Artificial Intelligence, Machine Learning, Data Science, Computer Science, or a related discipline, and have developed expertise in machine learning and AI research.
You have worked on modelling, forecasting, or complex data analytics challenges and are motivated to contribute to the exploration of innovative artificial intelligence architectures addressing energy-related challenges.
You have:
- Experience in Machine Learning, Deep Learning, or Artificial Intelligence;
- Strong proficiency in Python and a Deep Learning framework such as PyTorch or TensorFlow;
- Knowledge of predictive modelling, forecasting, or time-series analysis;
- Experience conducting research projects and scientific experimentation;
- Fluency in English, written and spoken (minimum C1 level) enabling you to work effectively in an international environment.
Experience in Computer Vision, Self-Supervised Learning (SSL), representation learning, or World Models and JEPA architectures would be considered a strong asset for this position.