Post-doc : Deep Learning for Predictive Mapping: Multi-Resolution Data Integration and Spatial Extrapolation

Contract duration: 12 months with possible extension to 24 months

This postdoctoral position aims to develop a methodology for the spatial extrapolation of subsurface attributes by integrating heterogeneous and multi-resolution data. The objective is to merge annotated logs and attribute maps while spatially constraining predictions to ensure their geographical consistency. The developed methodologies will be tested on both synthetic and real data for various use cases related to the energy transition (geothermal energy, geological CO2 sequestration, natural hydrogen, lithium extraction).
The PhD degree must have been obtained no more than three years before the start of the postdoctoral contract.

he objectives of this research proposal are threefold: design a methodology for merging heterogeneous data (logs, maps) ; develop predictive models to extrapolate a known attribute from logs to fine-scale grids covering the entire study area ; integrate graphs encoding the subsurface geometry to spatially constrain predictions and improve their consistency.

To achieve these objectives, the activities are structured around three main steps:

• Preparing and analyzing the different data sources: annotated logs, attribute maps, and graphs representing subsurface geometry
• Building a spatial extrapolation model based on heterogeneous data fusion—this will involve conducting a literature review
• Validating predictions and analyzing uncertainties.

The results will be disseminated through publications. The recruited postdoctoral researcher will work with teams comprising both data science specialists and geoscientists, who will provide domain-specific expertise on these case studies.

Compétences techniques

• Data science with strong proficiency in deep learning.
• Proficiency in Python programming and deep learning frameworks is essential.
• Knowledge of or interest in geosciences.
• Experience in graph learning or hybrid models is preferred.

Compétences comportementales

• Teamwork skills
•  Motivation for scientific research
• Autonomy
• Rigor
• Curiosity
  

The candidate must hold a PhD in data sciences, deep learning or statistics.

The candidate must also demonstrate excellent written and verbal communication skills in English. Knowledge of French is a plus.

For further information and to apply, please contact:

aurelie.chataignon@ifpen.fr