Research Fellow (DesCartes – WP2-8)

Job offer posted on 16 January 2023. Reposted on 19 October 2023. 


Research Fellow position in Faster than real-time Physics, addressing the issue of accuracy assessment and uncertainty quantification when using limited noisy data and approximate simulation tools (e.g. reduced order modeling) for fast data assimilation, prediction, and control.


CNRS@CREATE Ltd., the first CNRS’ overseas subsidiary, acts as a program operator to build and conduct large transdisciplinary research programs. It is a trans-continental hub for research with unique opportunities for high quality faculties, researchers & postgraduate/postdoctoral students. Located in Singapore, on the Campus for Research Excellence And Technological Enterprise (CREATE) it’s main purposes are: (i) To strengthen France and Singapore’s global position in the areas of research of the highest potential for present and future society; (ii) To create unique initiatives in trans disciplinary research of excellence and technology development which would not exist under normal circumstances or within a single country or institution; and (iii) To encourage direct industry participation in research projects to ensure efficient transformation of research results into innovative products.

Program DesCartes aims to develop disruptive hybrid AI to serve the smart city and enable optimized decision-making in complex situations for critical urban systems. Hybrid AI will help us support smart city critical infrastructures: (i) Smartly, with less data (knowledge / physics AI); (ii) Safely, certified and regulated (trustworthy AI); (iii) Carefully, in a human centric way (human AI); and (iv) Responsibly, by empowering people (societal AI).

Read more about the DesCartes program here.

The project is split into several work-packages (WPs), and among them:

  • WP2, called Learning from smart and complex data for hybrid AI, deals with the management of data which is hard to get, expensive, incomplete, or access-limited. The WP thus focuses on learning decision-making policies from hybrid twins in complex/uncertain settings, where taking into account physics-based knowledge can reduce the amount of data and computational resources needed.
  • WP8, called Augmented Hybrid Engineering, deals with practical aspects in terms of smart data acquisition (sensor placement, data fusion,…) and fast computations (model reduction, hybrid twin synthesis, robust control) for complex engineering systems.


In both WPs, fundamental research is conducted, developing methodologies which are agnostic with respect to any potential application or specific physics, even though specific case studies on systems of the smart city are targeted as proofs of concept. A common research topic of interest is the use and management of tools enabling faster than real-time physics description, connecting experimental data and simulation models, and beneficially used for data assimilation, prediction, and decision-making. This is the context of the collaboration between these 2 WPs.


Connecting data and models in a sound manner is not a trivial task. On the one hand, experimental information coming from data is sparse and noisy. On the other hand, models use to simulate complex systems (with strong nonlinearities, multiscale aspects, interactions between system components…) have limitations. These come from both model coarsening (e.g. use of approximate model reduction techniques such as POD, PGD or Reduced Basis method) and lack of knowledge on physical phenomena to be represented.

Therefore, a fundamental question addressed in the RF position is to efficiently relate the quality of data information with the accuracy of simulation models. In other words, the aim is to quantify, propagate, and link uncertainties coming from both worlds. This would enable to compute right and at right cost when performing fast data assimilation and decision-making.

In order to tackle this challenge, we will develop stochastic methodologies to inform on the quality of inputs (measurements and models) and propagate them to outputs through data assimilation and decision-making procedures. Uncertainties will be considered as small perturbations, and a specific focus on the representation, assessment and management of the quality of data and (reduced) models will be made; Suitable frameworks such as those related to Bayesian inference may be considered for this objective.

The developed methodologies will be first tested and validated on academic examples with synthetic noisy data and representative complex models, before being applied on some industrials cases considered in the DESCARTES programme.



Data assimilation

Numerical simulations


Model reduction

Model updating (inverse problems)

Bayesian inference / Kalman filtering

Uncertainty quantification/propagation


Applicants should have a background in applied mathematics for engineering, with interest in data analysis. They should hold a PhD in Applied Mathematics, Mechanical Engineering, or Computer Science. Specific kills in simulation-based engineering, machine learning, and programming (Python, C++)  will be appreciated.

It is expected that the candidate shows high motivation for the project, as well as  good communication (writing, oral) skills for publications in conferences or in scientific journals, and for collaborations with colleagues in the Descartes project. Ability to work in an international environment (with English working language), learning from experienced researchers and transfer knowledge, is also a mandatory requirement.


RF Salary range: 6000 to 7200 SGD (depending on suitability and experience)

Workplace address: CREATE Campus, CREATE Tower, 1 Create Way #08-01 Singapore 138602

Interested applicants should send a full CV (with attached motivation letter and contact information of two references)  to the following persons:


Ludovic CHAMOIN, Ecole Normale Supérieure (ENS) Paris-Saclay

Pierre SENELLART, Ecole Normale Supérieure (ENS)


Stéphane BRESSAN, National University of Singapore (NUS)



View this job offer on