Samy A. Foudil

Je conçois, déploie et industrialise des solutions intelligentes de traitement automatisé de documents complexes, en combinant Machine Learning, Deep Learning, modèles LLM et services AWS (Textract, Rekognition, SageMaker, etc.). Mon objectif : construire des workflows robustes, scalables et sécurisés pour extraire, enrichir et interpréter automatiquement l’information issue de documents non structurés. ◦ Conception et industrialisation de pipelines ML pour l’analyse automatique de documents (PDF, images, formulaires). ◦ Exploitation de LLM pour contextualiser, structurer et interpréter les résultats issus de Textract et Rekognition. ◦ Déploiement, automatisation et monitoring de modèles sur des architectures cloud scalables (AWS SageMaker). ◦ Mise en œuvre de mécanismes d’alerte et de supervision en production. ◦ Optimisation continue des performances, de la scalabilité et de la résilience des systèmes. ◦ Application des meilleures pratiques en matière de sécurité et conformité (RGPD, IAM, chiffrement). ◦ Conseil et accompagnement dans le choix et l’intégration de solutions IA adaptées au niveau d’expertise des équipes. ◦ Force de proposition dans la résolution de problématiques complexes liées à l’automatisation documentaire.

Design and implementation of an innovative system combining web and mobile applications with APIs for the collection, processing, and analysis of geolocated behavioral data. The project is part of an experimental cognitive research approach using virtual environments and real-life contexts to study episodic memory. ◦ Contribution to the study of episodic memory through technological tools enabling the measurement of the subjective quality of memory in relation to the spatio-temporal encoding context. ◦ Integration of AI models for the extraction of insights from contextual data (location, time, action) collected in naturalistic conditions, in line with the work of Foudil et al. (2021, 2024). ◦ Full-stack development of a web application, REST APIs, and a mobile app, enabling real-time data analysis and interactive visualization through analytical dashboards. ◦ Implementation of automated CI/CD pipelines (Docker, GitHub Actions), ensuring smooth deployment of the software infrastructure. ◦ Software architecture consulting to ensure scalability, sensitive data security, and compliance with ethical research standards.

Management of web application and data infrastructure development. Management of the BrainWatch BrainWatch research project (Deep Tech, CIR). ◦ Development of BrainWatch behavioral collection and analysis platform. ◦ Design and development of systems infrastructure and mobile application. ◦ Implementation of deployment pipelines and RGPD compliance. ◦ Management of development teams and design of dedicated mobile application. ◦ R&D leadership focused on the collection and analysis of large amounts of behavioral data, as well as the design of an intelligent investigation system using GenAI.

Design and analysis of fMRI studies on social interactions between individuals. The aim was to propose a complete experience acquisition and analysis pipeline optimized for fMRI studies of social interactions between individuals. ◦ BIDS format management, SPM12 preprocessing + 1fst & 2nd level, LSA and LSS modeling ◦ CoSMo MVPA: searchlight + ROI analysis, SVM ◦ Analysis of millions of voxels for fMRI and behavioral studies. ◦ Design of a dedicated PostgreSQL database to facilitate processing.

The aim of my thesis is to develop a new research paradigm aimed at better understanding human cognitive functioning, in particular episodic memory, in people's everyday lives. The experiments conducted will make innovative use of web and virtual reality technologies to observe and measure participants' behaviors in both real and virtual situations. This will enable us to quantify and model the complexity of episodic memory and delineate the underlying cortical structures. ◦ As part of my thesis in cognitive neuroscience, I designed a complete computational ecosystem to study human memory. ◦ Developed computerized processes for collecting behavioral data generated in participants' daily lives, as well as using functional magnetic resonance imaging. ◦ Implementation of a hybrid computing infrastructure using laboratory resources and the AWS cloud for computation. ◦ Creation of a dedicated mobile application, integrated with a suite of specially developed APIs to facilitate data collection and analysis. ◦ Developed innovative machine learning algorithms to address my research hypotheses. ◦ Web administration interface and dashboard for visualization and control of experiments.

Processing of a large raw data set. Temporal analysis and visualization of this data on the Orange-Undertracks platform. Data from the Grenoble Faculty of Medicine. This work is part of the ANR Hubble project. One of my objectives was to determine the behavioral typologies of students according to the pedagogies applied. ◦ Data collection and behavioral analysis to cconnaitre the effects of pedagogical strategies. ◦ Design of dashboards to visualize results. ◦ Design of a dedicated MySQL database.

Setting up tracings to analyze the effectiveness of personalization strategies as part of the EDUCLEVER “Cartographie des Savoirs” project As part of the “Cartographie des Savoirs” project, an initial personalization model has been set up. This model is based on the definition of pedagogical strategies for proposing remediation sessions to learners. My task is to define how to trace the personalization engine and to implement this tracing so as to be able to present initial results on the effectiveness of the strategies implemented, and of the engine executing these strategies. To do this, I need to identify the elements to be traced on the engine side, to see how the strategies are used, and on the learner side, to know what the learner is doing, and how to trace the impact of the proposed remediation.

Evaluation de l'appropriation d'un serious game : Tamagocours. Création d'un modèle d'appropriation selon une approche pluridisciplinaire (cognitive, sociologique, ergonomique) et basé sur l'analyse des traces informatisées des utilisateurs. Création d'une méthode de traitement automatique des données (Machine Learning, Clustering) dans le but de visualiser l'état d'appropriation du dispositif par groupe d'utilisateurs et par utilisateur. Sous la direction d'Eric Sanchez, Maître de conférences, Didactique des sciences et TICE, EducTice- S2HEP, Institut français de l'éducation, École Normale Supérieure de Lyon.

Validation of an innovative system for recognizing young children's oral productions (in French), providing a methodological tool for researchers and practitioners specializing in language acquisition. Under the supervision of Mélanie Canault (Senior Lecturer) and Hung Thai-Van (University Professor and hospital practitioner) at the DDL Laboratory (UMR 5596, CNRS, Université Lyon 2) and Audiology and Orofacial Explorations Department, Hospices Civils de Lyon & Lyon Neuroscience Research Center, INSERM U1028-CNRS UMR 5292.

Maintenance et administration du réseau informatique et télécom. Déploiement d'une solution IPBX (Asterisk) pour l'ensemble des agences de la CPAM.