Talks
My conference presentations and poster sessions at academic venues.
Conference Presentations
Michelin DoctoBib'Day 2026
A Unified Online Framework for Adaptive Soft Sensing in High-Dimensional Batch Processes
April 13, 2026 — Clermont-Ferrand, France
Louis Berthier1,2, Ahmed Shokry1, Maxime Moreaud2, Guillaume Ramelet2, Eric Moulines1
1 CMAP, Ecole Polytechnique 2 Michelin, Clermont-Ferrand
How do you predict product quality in real time when the manufacturing process itself keeps changing? This work benchmarks five adaptive soft sensing strategies, from temporal weighting to neighborhood-based retrieval, leveraging self-organizings with online SHAP-driven feature selection. Evaluated on over 35,000 production batches and 167 process variables at Michelin, it provides practical guidance for choosing the right online adaptation strategy in industrial settings.
ESCAPE 35
Knowledge Discovery in Large-Scale Batch Processes through Explainable Boosted Models and Uncertainty Quantification: Application to Rubber Mixing
July 9, 2025 — Ghent, Belgium
Louis Berthier1,2, Ahmed Shokry1,*, Eric Moulines1, Sylvain Desroziers1, Guillaume Ramelet2
1 CMAP, CNRS, Ecole Polytechnique, IP Paris 2 Michelin
Rubber compounding involves hundreds of interacting process variables, but which ones actually drive product quality? This work takes an explainability-first approach, combining gradient boosted trees with SHAP attribution and conformal prediction to give engineers both interpretable insights and statistically rigorous uncertainty estimates. The result: process experts can pinpoint critical quality drivers with quantified confidence. Presented at ESCAPE 35, one of Europe's premier conferences in computer-aided chemical engineering.
Posters
Local Dynamic Calibration via JiT-CP
April 13, 2026 — Clermont-Ferrand, France — Michelin DoctoBib'Day 2026
Louis Berthier, Ahmed Shokry, Maxime Moreaud, Guillaume Ramelet, Aymeric Dieuleveut
Standard conformal prediction gives you prediction intervals that are correct *on average*, but can be dangerously wrong for specific process conditions. JiT-CP fixes this by retrieving the most similar historical batches through SOM-based similarity search and computing locally weighted conformal scores, producing prediction intervals that adapt to whichever regime the process is currently in.
Local Dynamic Calibration via JiT-CP
March 22-25, 2026 — Minneapolis, USA — SIAM UQ26
Louis Berthier, Ahmed Shokry, Maxime Moreaud, Guillaume Ramelet, Aymeric Dieuleveut
Presented at SIAM UQ26, a leading conference on uncertainty quantification dedicated to industrial applications. JiT-CP tackles a well-known blind spot of conformal prediction: marginal coverage guarantees can hide systematic failures for specific subpopulations. By coupling Just-in-Time Learning with a SOM-based latent space, the method delivers locally adaptive prediction intervals with group-conditional coverage, no distributional assumptions, no retraining required.
Knowledge Discovery in Large-Scale Batch Processes
December 2024 — Paris, France — Welcome Day IP Paris
Louis Berthier, Ahmed Shokry, Eric Moulines, Sylvain Desroziers, Guillaume Ramelet
How do you turn a black-box quality prediction into something a process engineer can act on? This poster presents a unified framework pairing gradient boosted models with SHAP attribution and conformal coverage guarantees, translating raw predictions into interpretable, uncertainty-aware process insights at production scale. Awarded Best Poster in the Mathematics category.
A Framework for Knowledge Discovery in Rubber Mixing Processes
November 2024 — Clermont-Ferrand, France — Michelin Doctoral Day
Louis Berthier, Ahmed Shokry, Eric Moulines, Sylvain Desroziers, Guillaume Ramelet
The first iteration of my knowledge discovery framework, focused on offline analysis. Gradient boosted regression combined with SHAP attribution and conformal prediction intervals surfaces the most influential process variables and their interactions, giving rubber compounding engineers a clear, quantified view of what drives product quality.
2DSBG: A 2D Semi Bi-Gaussian Filter for Line Feature Detection
June 2023 — Rhodes, Greece — ICASSP 2023
Louis Berthier, Adrien Ruggiero, Marcel Pie, Ghulam Sakhi Shokouh, Baptiste Magnier
How do you detect thin line features in noisy images with sub-pixel accuracy? The 2DSBG filter uses an asymmetric Gaussian kernel that selectively enhances elongated structures while suppressing background noise, outperforming classical symmetric approaches like the Laplacian-of-Gaussian on both synthetic and real-world benchmarks. Presented at ICASSP, the flagship IEEE conference on signal processing.