Thibault Nadin Email & Phone Number

Ville De Paris, Île-De-France, France

Paris Digital Lab is an open innovation training program that enables students to acquire skills related to many different technologies, but also to meet the expectations of real-world businesses.𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 - 𝗥𝗲𝗽𝗮𝗶𝗿 𝗪𝗮𝗿𝗿𝗮𝗻𝘁𝘆 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 (𝗡𝗲𝘀𝗽𝗿𝗲𝘀𝘀𝗼)- Performed text clustering and model fine-tuning to determine warranty coverage- Implemented a topic recognition system for technical team intervention reports. By fine-tuning GPT models to… Show more Paris Digital Lab is an open innovation training program that enables students to acquire skills related to many different technologies, but also to meet the expectations of real-world businesses.𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 - 𝗥𝗲𝗽𝗮𝗶𝗿 𝗪𝗮𝗿𝗿𝗮𝗻𝘁𝘆 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 (𝗡𝗲𝘀𝗽𝗿𝗲𝘀𝘀𝗼)- Performed text clustering and model fine-tuning to determine warranty coverage- Implemented a topic recognition system for technical team intervention reports. By fine-tuning GPT models to embed free-text entries.- Achieved 80% automation in Nespresso's warranty processing with an error rate under 3%.⚙️ PyTorch, Transformers, Azure𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 - 𝗣𝗮𝗿𝘁𝗶𝗮𝗹 𝗗𝗶𝘀𝗰𝗵𝗮𝗿𝗴𝗲 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 (𝗡𝗲𝘅𝗮𝗻𝘀) - Developed a new approach by combining convolutional networks and statistical learning to detect partial discharges.- Localized premature wear of electrical cables on the electric grid.- Deployed the model, enabling real-time detection and localization of faults.⚙️ Scikit-learn, Pandas, PyTorch, Git, Jupyter𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 - 𝗔𝗜 𝗕𝗮𝘀𝗲𝗱 𝗠𝗖𝗤𝘀 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗼𝗿 (𝟮𝟬/𝟮𝟬 𝗖𝗼𝗻𝗰𝗼𝘂𝗿𝘀 𝗲𝘁 𝗘𝘅𝗮𝗺𝗲𝗻𝘀)- Developed a collaborative application prototype for creating multiple-choice questionnaires.- Introduction of generative AI tools in the web application.⚙️ PyTorch, Azure, React, Django, Git Show less

𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗗𝗲𝗲𝗽𝗳𝗮𝗸𝗲 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗖𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 (𝟭𝘀𝘁 𝗽𝗹𝗮𝗰𝗲)- 24-hour competition involving over 15 leading engineering schools, tasked with building a model proficient in detecting video-based deepfakes.- Used computer vision tools to detect faces- Trained a 3D-CNN to perform classification.⚙️ PyTorch · Computer Vision · Deep learning𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗖𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 (𝗞𝗮𝗴𝗴𝗹𝗲, 𝘁𝗼𝗽𝟱%) -… Show more 𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗗𝗲𝗲𝗽𝗳𝗮𝗸𝗲 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗖𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 (𝟭𝘀𝘁 𝗽𝗹𝗮𝗰𝗲)- 24-hour competition involving over 15 leading engineering schools, tasked with building a model proficient in detecting video-based deepfakes.- Used computer vision tools to detect faces- Trained a 3D-CNN to perform classification.⚙️ PyTorch · Computer Vision · Deep learning𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗖𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝗼𝗻 (𝗞𝗮𝗴𝗴𝗹𝗲, 𝘁𝗼𝗽𝟱%) - Extracted features from satellite observations to achieve classification of buildings - Leveraged key features and constructed appropriate machine learning models. - Used semi-supervised learning to deal with the unlabeled data available.⚙️ Python : PyTorch · Scikit-learn · Pandas · Feature Engineering Show less

𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗜𝗺𝗮𝗴𝗲 𝗦𝘂𝗽𝗲𝗿-𝗥𝗲𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻 - Developed network-based learning algorithms aimed at restoring the quality of degraded images. - Engineered a model tailored for specific applications such as face upscaling, optimized for low memory usage and rapid inference.⚙️ TensorFlow · PyTorch · Image Processing · NumPy · Deep learning · Signal Processing𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 - 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻 𝗼𝗳 𝗡𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗘𝗻𝗲𝗿𝗴𝘆 𝗗𝗲𝗺𝗮𝗻𝗱… Show more 𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 - 𝗜𝗺𝗮𝗴𝗲 𝗦𝘂𝗽𝗲𝗿-𝗥𝗲𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻 - Developed network-based learning algorithms aimed at restoring the quality of degraded images. - Engineered a model tailored for specific applications such as face upscaling, optimized for low memory usage and rapid inference.⚙️ TensorFlow · PyTorch · Image Processing · NumPy · Deep learning · Signal Processing𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 - 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻 𝗼𝗳 𝗡𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗘𝗻𝗲𝗿𝗴𝘆 𝗗𝗲𝗺𝗮𝗻𝗱 (𝗖𝗮𝗽𝗴𝗲𝗺𝗶𝗻𝗶 𝗜𝗻𝘃𝗲𝗻𝘁) - Built an accurate predictive model for electricity consumption. - Compared the performances of classical ML models (ARIMA, Random Forest, KNN, XGBoost) andneural-net-based models (MLP, RNN, LSTM). - Explored feature engineering and feature selection techniques to achieve better results.⚙️ PyTorch · Scikit-learn · Statistical Learning · NumPy · Time series analysis · Pandas𝗣𝘆𝘁𝗵𝗼𝗻 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 - 𝗚𝗲𝗻𝗲𝘁𝗶𝗰 𝗔𝗹𝗴𝗼𝗿𝗶𝘁𝗵𝗺Led the development of a genetic algorithm to address the DNA plasmid folding problem, building on concepts from a game theory course.⚙️ Docker · Python𝗣𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗶𝗻𝗴 - 𝗖𝗼𝗻𝗻𝗲𝗰𝘁𝟰Developed the Connect Four game in Python using Pygame, incorporating AlphaBeta algorithm and reinforcement learning techniques to design an artificial intelligence system.⚙️ Artificial intelligence · Computational Game Theory · Reinforcement Learning · Python Show less

Automatants is the Artificial Intelligence association of CentraleSupélec. Its purpose is to enable students to become proficient in AI by providing opportunities to engage in hands-on projects, and interact with experts in the field.𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗘𝘃𝗲𝗻𝘁𝘀 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁:- Provided information to 3000 campus students about the association's initiatives.- Promoted the association's activities on social media platforms.- Orchestrated "la Soirée de… Show more Automatants is the Artificial Intelligence association of CentraleSupélec. Its purpose is to enable students to become proficient in AI by providing opportunities to engage in hands-on projects, and interact with experts in the field.𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗘𝘃𝗲𝗻𝘁𝘀 𝗺𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁:- Provided information to 3000 campus students about the association's initiatives.- Promoted the association's activities on social media platforms.- Orchestrated "la Soirée de l'IA", a series of three mini-conferences (talks) that showcased research, innovation, and key inquiries related to artificial intelligence for an audience of 100 people.𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗼𝗿 𝗶𝗻 𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗮𝗻𝗱 𝗜𝗺𝗮𝗴𝗲 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴:- Spearheaded instruction in Deep Learning and image processing, focusing on Neural Networks programming using PyTorch and TensorFlow.- Led a dynamic team of 10 students as the primary instructor for a cutting-edge Neural Style Transfer project, fostering a collaborative environment and ensuring successful project completion.- Delivered comprehensive lectures and hands-on workshops, equipping students with practical skills in PyTorch and TensorFlow for Neural Network development. Show less

Production of Automotive Injector Prototypes on the Pilot Line of the Factory and Production Control (Team of 40 People).

Digital Tech Year

The Digital Tech Year is a selective computer science track, made up of 6 months of agile prototyping at the Paris Digital Lab.

Master Of Engineering, Applied Mathematics And Computer Science

GPA: 4.10/4.33 Main courses (1A) : - Measure Theory and Probability - Algorithmics & Complexity - Computational Game Theory - Partial.

Classes Préparatoires Aux Grandes Écoles (Cpge)

GPA: 3.99/4 Mathématics, Physics, Informatics. Préparation for top engineering schools' entrance exams.