Conceptualisation and design of a DNN inference framework dedicated to massive parallel simulation on exascale architectures

Status

Filled

Scientific disciplines

Mathematics

Research direction

Digital Science and Technology

Affiliate site

Rueil-Malmaison

INRIA and IFP Energies nouvelles, leading French research laboratories, are seeking a talented and motivated master's student to join their team for a PhD position. As numerical simulation plays a pivotal role in our research and industrial applications, optimizing the performance of our parallel simulators is crucial. Deep learning (DL) methods have shown great promise in enhancing traditional numerical methods, and their application in high performance computing (HPC) is on the rise.
The objective of this PhD research is to develop a methodology for coupling massively parallel applications with inference engines in the context of exascale machines. The successful candidate will work on conceptualizing and designing a deep neural network (DNN) inference framework tailored specifically for massive parallel simulations on exascale architectures.
As a PhD candidate, you will:
•    Investigate state-of-the-art DL algorithms and parallel computing techniques to develop an understanding of their applicability in the context of exascale simulations;
•    Conceptualize and design a DNN inference framework that seamlessly integrates with massively parallel simulation codes developed in low-level languages such as C, C++, or Fortran;
•    Implement the proposed framework and optimize its performance to leverage the capabilities of exascale architectures effectively;
•    Conduct rigorous testing and validation of the framework using real-world simulation scenarios to ensure reliability and accuracy.
The PhD candidate will join a dynamic and collaborative research environment at INRIA and IFP Energies nouvelles. He/she will have the opportunity to work on cutting-edge research at the intersection of DL and HPC and will benefit of mentorship from leading experts in the field, competitive stipend and research funding.

Keywords: Machine-Learning, Deep-Learning, CFD, HPC, GNN, Data science, …

  • Academic supervisor    Prof. Bruno RAFFIN, bruno.raffin@inria.fr DataMove - INRIA Grenoble
  • Doctoral School    ED217 MSTII, Université Grenoble Alpes
  • IFPEN supervisors    Dr Jean-Marc GRATIEN  & Dr Raphael GAYNO
  • PhD location    INRIA, DataMove, Grenoble, France
  • Duration and start date    3 years, starting in September 2024
  • Employer    INRIA 
  • Academic requirements    Master's degree in Computer Science, Engineering, Scientific Computing, Data Science or a related field. Strong background in parallel computing, HPC, and Deep Learning. Proficiency in programming languages such as Python, C, C++, or Fortran.

How to Apply:
Please send your cover letter and CV to the supervisors indicated below.

Contact
Encadrants IFPEN :
Dr Jean-Marc GRATIEN & Dr Raphael GAYNO