Informations générales
Intitulé de l'offre : PhD Position (M/F) (H/F)
Référence : UMR7030-HANAZZ-002
Nombre de Postes : 1
Lieu de travail : VILLETANEUSE
Date de publication : lundi 7 avril 2025
Type de contrat : CDD Doctorant
Durée du contrat : 36 mois
Date de début de la thèse : 1 octobre 2025
Quotité de travail : Complet
Rémunération : 2200 gross monthly
Section(s) CN : 06 - Sciences de l'information : fondements de l'informatique, calculs, algorithmes, représentations, exploitations
Description du sujet de thèse
The objective of this thesis is to propose a digital twin patient model designed to (i) generate individualized SEPSIS patient profiles and (ii) predict, at early stages of the pathology, patient-specific responsiveness to corticosteroids based on the most relevant biomarkers. To generate these SEPSIS profiles, we adopt a strategy based on advanced deep generative learning models. To estimate corticosteroid responsiveness, we propose leveraging reinforcement learning from human feedback (RLHF). This approach is particularly promising for the design of personalized and adaptive treatment strategies.
RLHF will enable training of large language models (LLMs) by incorporating knowledge and insights from medical experts, modeled through a knowledge graph. This thesis focuses on four main challenges:
(1) Multimodal data fusion, integrating heterogeneous medical data sources (clinical, biological, imaging, etc.). This remains a major challenge for LLMs, which are primarily trained on text and image data.
(2) The creation of personalized patient profiles and prediction of corticosteroid responsiveness.
(3) Evaluation and monitoring of corticosteroid efficacy across different stages of the disease's progression.
(4) Implementation of causal inference methods to identify the most effective treatment strategies that minimize SEPSIS-related complications.
By combining data-driven models with medical expertise within a unified cognitive framework, this work aims to contribute significantly to the development of patient-centered digital health solutions.
Contexte de travail
The LIPN laboratory, particularly the A3 team, has developed original algorithms in the field of machine learning and artificial intelligence for mining large volumes of data with varying natures and structures. Most of the research work is driven by both external and internal collaborations within the university, notably through ANR and CIFRE projects.
Contraintes et risques
No
Informations complémentaires
supported by 80PRIME 2025