About | Victor Bertret

Who I am

I am an AI & Control Engineer at Purecontrol, where I design machine learning and stochastic control algorithms to optimize the operation of real industrial systems, cutting energy consumption while respecting strict operational and environmental constraints.

I hold a PhD in Applied Mathematics from the Université de Rennes, defended in December 2025. My work sits at the intersection of physical modeling, statistics / data assimilation, and optimal control, with a strong focus on bringing these methods all the way to industrial deployment.

My current work is mainly in the water sector, and industry more broadly: the estimation and control problems these systems raise, recent optimization methods, and how to formulate those problems for real, non-stationary installations that are hard to characterize.

Main research interests

Stochastic control: robust decision-making under uncertainty (stochastic dynamic programming, MPC).
Data assimilation: real-time state estimation for non-linear, partially observed systems (Kalman & particle filters).
Industrial optimization: energy efficiency in water and biogas processes (MILP, optimal control).
Hybrid modeling: combining physical models (ODEs / ASM1) with data-driven and machine learning components.

PhD in a nutshell

My thesis, Machine Learning and Stochastic Control for the Optimized Automatic Piloting of Industrial Systems, was carried out under a CIFRE fellowship between Purecontrol, the IRMAR mathematics laboratory, and the IETR / ENS Rennes. It develops a unified methodology to optimize the aeration of wastewater treatment plants under discharge constraints and strong uncertainty, combining the identification of stochastic state-space models with stochastic optimal control.

Supervised by Valérie Monbet (director, Université de Rennes), Roman Le Goff Latimier (ENS Rennes) and Gautier Avril (Purecontrol). → Read more

Selected publications

Data assimilation for prediction of ammonium in WWTP: from physical to data driven models
Water Research, Vol. 282 (2025), journal article · details
Predictive control for sensorless phosphorus removal
Congrès Eau & IA 2026, Grenoble, with Veolia · details
A stochastic EM algorithm for the estimation of WWTP ammonium concentration
European Control Conference (ECC 2024), Stockholm · details

See all research & talks →

Supervision

Farah Dogui
Predictive modelling of biogas production in anaerobic digestion: forecasting output from the feedstock and operating conditions to optimise plant yield.
Engineering internship (M1, IMT Nord Europe), Purecontrol, 2026
Mathias Lommel
Data-driven strategies to speed up a mixed-integer linear programming (MILP) solver.
Final-year engineering project (PFE, INSA Rennes), Purecontrol, 2025
Nay Klaimi
Data-driven dynamic modelling, applied to a wastewater treatment plant.
Master's research internship (PFE, INSA Rennes), Purecontrol, 2024

Technical toolbox

Julia Python Stochastic control Model Predictive Control Data assimilation Kalman / particle filters MILP (HiGHS, SCIP) State-space modeling Machine learning

Connect

Based in Rennes, France. Happy to talk about industrial innovation, applied mathematics, or environmental challenges: victor.bertret [at] purecontrol.com.