Compositional Hybrid Digital Twins Models

Engineering Digital Twins (EDT); France; Ibrido; EUR 20.000 - 27.000; Tempo pieno

Vantaggi offerti da questo lavoro

3-year duration

Competitive PhD stipend

Social security & health coverage

Research travel support

Descrizione del lavoro

Engineering Digital Twins (EDT) invites applications for a PhD candidate to develop hybrid digital twin frameworks for fermentation ecosystems, focusing on compositional modeling operators and reusable hybridization patterns.

The project combines mechanistic ecology with machine learning and utilizes experimental datasets from vegetable fermentation to calibrate and validate models within the EDT program.

Competenze

• Master’s degree in Computer Science, Computational Biology, Applied Mathematics, or related field.
• Strong interest in modeling and analyzing complex biological or dynamical systems.
• Programming skills in scientific computing environments (e.g., Python, Julia, or similar).

Mansioni

• Conduct research on compositional hybrid modeling frameworks for digital twins applied to microbial ecosystem modeling.
• Design hybridization operators integrating mechanistic and data-driven models.
• Develop reusable hybrid modeling patterns for microbial ecosystem simulation.
• Develop and implement mechanistic models of microbial fermentation dynamics (e.g., consumer–resource models).
• Implement prototype digital twin architectures for fermentation systems.
• Develop methods for parameter inference and model calibration from fermentation time-series data.
• Publish research findings in leading conferences and journals.
• Participate in interdisciplinary collaboration within the EDT program.

Conoscenze

Python

Machine Learning

Dynamical Systems

English communication

Formazione

Master's degree in Computer Science, Computational Biology, Applied Mathematics, or related field

Strumenti

Python

Julia

Descrizione del lavoro

We are seeking a motivated PhD candidate to contribute to the Engineering Digital Twins (EDT) program within Catalyst: the Reliable Hybrid Model Forge. The research focuses on developing hybrid digital twin frameworks, with an emphasis on compositional hybrid modeling operators and reusable hybridization patterns.

Fermented food systems are complex microbial ecosystems shaped by dynamic interactions between species and their environment. Predictive and controllable models of these systems are essential for advancing food innovation, safety, and functionality. However, these ecosystems involve highly nonlinear, evolving interactions that are difficult to capture using a single modeling paradigm.

This project proposes to experiment through the development of a digital twin for fermented microbial communities, grounded in mechanistic ecological models and enriched with machine learning components. The central goal is to design composable hybrid modeling operators that enable systematic integration of mechanistic and data-driven models in a reusable and scalable digital twin architecture.

Research Focus

Microbial fermentation systems involve complex interactions between microorganisms, metabolites, and environmental conditions. Mechanistic frameworks such as the consumer–resource model introduced by Niehaus et al. (2019) provide a strong theoretical basis for modeling these systems, capturing resource-mediated interactions such as competition, facilitation, and self-restraint.

However, purely mechanistic approaches may become computationally expensive or insufficient when dealing with high-dimensional experimental data or partially understood processes. Hybrid modeling approaches that combine mechanistic models with machine learning–based surrogate models provide a promising alternative.

This research will investigate how such hybrid models can be constructed through formal operators of model composition, enabling systematic hybridization across modeling paradigms.

The PhD will explore the following research directions:

Hybridization Operators for Microbial Digital Twins
Designing operators that combine mechanistic ecological models with data-driven components in a consistent simulation framework, for example by learning corrections to partially specified dynamics or by inferring latent processes from incomplete observations.

Reusable Hybrid Modeling Patterns
Defining compositional patterns that allow recurring hybridization strategies (e.g., surrogate replacement, model augmentation, residual learning, multi-fidelity switching).

Multi-Fidelity Model Composition
Supporting adaptive switching or blending between models of different complexity, such as consumer–resource models, generalized Lotka–Volterra models, replicator dynamics, and machine learning surrogates.

Metadata and Interface Standardization
Developing standardized interfaces and metadata structures enabling consistent composition of heterogeneous models within digital twin workflows.

These approaches aim to enable flexible digital twin architectures capable of handling incomplete observations and evolving biological knowledge, where hybrid models can be composed, replaced, or extended using well-defined hybridization operators.

Data and Experimental Context

The proposed framework will be instantiated using experimental datasets from vegetable fermentation studies, including data generated within the FermenTwin project as well as datasets available in the fermentation literature.

These datasets include fermentation experiments performed on various vegetable substrates, allowing the study of microbial ecosystem dynamics under diverse ecological conditions.Fermentation time series typically contain longitudinal observations of microbial community composition and metabolite dynamics across different stages of fermentation, providing insight into microbial interactions and ecosystem evolution.

Additional datasets reported in the literature on vegetable fermentation microbial ecology (e.g., Wuyts et al., 2018) may also be used to benchmark and validate the proposed models.

Together, these datasets provide a ideal basis for calibrating, validating, and stress-testing hybrid digital twin models of microbial fermentation ecosystems across heterogeneous experimental conditions.

Key Responsibilities

• Conduct research on compositional hybrid modeling frameworks for digital twins applied to microbial ecosystem modeling
• Design hybridization operators integrating mechanistic and data-driven models
• Develop reusable hybrid modeling patterns for microbial ecosystem simulation
• Develop and implement mechanistic models of microbial fermentation dynamics (e.g., consumer–resource models)
• Implement prototype digital twin architectures for fermentation systems
• Develop methods for parameter inference and model calibration from fermentation time-series data
• Publish research findings in leading conferences and journals
• Participate in interdisciplinary collaboration within the EDT program

Research Environment

The PhD will be conducted within the Engineering Digital Twins (EDT) research program, which brings together researchers in digital twin engineering, modeling, and data science.

The project will combine expertise in:

• microbial ecology and fermentation science
• ecological and dynamical systems modeling of microbial ecosystems
• digital twin architectures
• hybrid modeling and simulation
• machine learning for scientific systems

Facilities and Resources

• Access to experimental fermentation datasets
• High-performance computing resources
• Collaboration with interdisciplinary researchers in biology, modeling, and digital twin engineering
• Opportunities for international collaborations

Qualifications

Required

• Master’s degree in Computer Science, Computational Biology, Applied Mathematics, or related field
• Strong interest in modeling and analyzing complex biological or dynamical systems
• Programming skills in scientific computing environments (e.g., Python, Julia, or similar)
• Background in machine learning and dynamical systems (ODE-based models))
• Strong analytical and problem-solving abilities
• Good communication skills in English
• Interest in interdisciplinary research at the interface of biology, modeling, and artificial intelligence

Preferred

• Knowledge of ecological or biological system modeling
• Experience with hybrid modeling or scientific machine learning
• Familiarity with digital twin concepts
• Experience with experimental datasets

Funding and Benefits

• Duration: 3 years
• Salary: Competitive PhD stipend according to French standards
• Benefits: Social security, health coverage, research travel support
• Travel: Support for conference attendance and research collaborations

Contact Information

For scientific questions regarding the PhD topic, please contact the main supervisor directly.

About the EDT Program

The Engineering Digital Twins (EDT) program is a major French research initiative dedicated to advancing the science and engineering of digital twin technologies through interdisciplinary collaboration.