Clement Wang

PhD Student in Graph Representation Learning

clementwang.pro@gmail.com

clementw168

clem-wang

https://clementw168.github.io/

Education

CIFRE PhD in Applied Mathematics

2025 - Present (Ongoing)

Institut Polytechnique de Paris - Mirakl

Graph Representation Learning with Self-Supervised Learning

Supervised by Prof. Thomas Bonald

MSc in Applied Mathematics - MVA

Sep 2023 - Dec 2024 (1 year)

École Normale Supérieure Paris-Saclay

Relevant coursework: NLP, graph learning, time series and signals, computer vision, geometric data analysis, optimization.

A few examples of project:

Master Thesis internship on Time Series self-supervised learning 🔗
Text-based Molecule Retrieval 🔗
Mixture Density Networks 🔗
Stochastic Weight Averaging (SWA) 🔗
Lymphocytosis Classification 🔗
Adaptive Brownian Bridge-based Symbolic Representation of Time Series 🔗

MSc in Data Science

Sep 2020 - Dec 2024 (4 years)

CentraleSupélec - Université Paris-Saclay

Relevant coursework: machine learning, statistics, stochastic calculus, parallel calculus, algorithmics, signal processing.

A few examples of project:

Multi-Agent Reinforcement Learning Project with Google DeepMind 🔗
Prey Predator Simulation with MADDPG 🔗
Aspect-Based Sentiment Analysis 🔗
Graph Neural Network Benchmark on Twitch Dataset 🔗
Reinforcement Learning applied to Flappy Bird 🔗
Implicit Acceleration by Overparameterization 🔗
Spiking Neural Networks 🔗

Preparatory Courses - MPSI/MP*

Sep 2018 - Jul 2020 (2 years)

Lycée Henri IV

Professional Experience

Data Scientist

Dec 2024 - Jul 2025 (9 months)

Mirakl • Permanent Paris, France

Built end-to-end delivery date estimation model for marketplace orders, from POC to production deployment

Serving 250,000+ orders per month
Increasing on-time delivery rate from 50% to 85%, reducing order late deliveries by 50%
Successfully launched beta program in July 2025 and general release in October 2025

Machine Learning Engineer

May 2024 - Oct 2024 (6 months)

Beacon Biosignals • Internship Paris, France

Trained a foundation model for EEG signals.

Adapting transformer tokenization for time series
25M parameters model trained on 2B tokens with contrastive learning and DINO teacher-student architecture
80.0% accuracy on sleep staging with a limited amount of annotated data, a 2.6% decrease with 95% fewer labels compared to fully supervised models

3D Computer Vision Engineer

Feb 2023 - Jul 2023 (6 months)

Stryker • Internship Freiburg-im-Breisgau, Germany

3D calibration of surgery instruments: Developing an end-to-end pipeline to determine the 3D geometry of a surgery instrument from a single RGB camera

2D key points detection with High resolution network (CNN)
3D triangulation from several video frames
Condensing the process from 2 minutes to 10 seconds with a 3.0 mm error

Co-Founder and MLOps Engineer

Dec 2022 - Jun 2023 (7 months)

Photogen AI • Co-founder Paris, France (Remote)

Founding a startup on personalizing diffusion models

Fine-tuning Stable Diffusion 1.5 with DreamBooth for realistic images on 10 images of a customer
Setting up cost-optimized cloud infrastructure on AWS with queue service, batch jobs, and autoscaling

Data Scientist Intern

Aug 2022 - Jan 2023 (6 months)

Polygon Technologies • Internship Santa Monica, CA, USA

Extraction of features from videos of cognitive tests to diagnose learning differences

Benchmarking Speech-to-text models (Whisper, Amazon Transcribe, Google Speech-to-Text)
Human pose detection, Face landmarks detection, Voice activity detection
Setting up a scalable pipeline to process 3-hour long videos

Machine Learning Consultant

Aug 2022 - Jan 2023 (6 months)

Paris Digital Lab • Consulting Paris, France

Various Computer Vision Projects for companies

Radio wave detection for the French Ministry of Armed Forces: Achieved 0.95 mAP @ IOU 0.5 with YOLOv3
Lipstick retrieval from selfies for L'Oréal Research & Innovation
Few shot object detection for Oorion: Annotation of object detection dataset with CLIP embeddings

Presidency of Automatants

2021 - 2022 (1 year)

Automatants, the AI student association of CentraleSupélec

Led a team of 20 members, organized events and courses, and promoted machine learning at CentraleSupélec, summarized 100+ papers

Linked articles:

My life as President 🔗
Winning an unbalanced image classification competition 🔗
Neural style transfer 🔗
Solving a maze with a genetic algorithm 🔗
My first Deep Learning project on GANs 🔗

Multi-Agent Reinforcement Learning Project

2023 - 2024 (1 year)

Google DeepMind

Accelerating the convergence of FoReL for Nash equilibrium finding in two-player zero-sum games.

Leaf Area Index prediction from satellite images Competition

2023 (3 months)

Won the first prize on this student competition over 30 participants. We published a paper on our solution at Big Data From Space 2023.

3D horror game on Unity for a one-week Game Jam

2021 (1 week)

CentraleSupélec

Developed a 3D horror game on Unity for a one-week Game Jam in a team of 5 students. We all learned C# one week before the game jam.

Machine Learning

Deep Learning
Graph Neural Networks, Time Series
Computer Vision, Natural Language Processing
Numpy, Pandas, PyTorch, TensorFlow, Scikit-learn, Hugging Face

Technologies

Python, C++, C#, C
Bash, Linux, Git, Docker, AWS
R, SQL, Spark
Jupyter, Airflow, MLflow, Weights & Biases
LaTeX

Languages

French - Native

English - Fluent

Mandarin - Conversational

Resume