Notes
Everything written up here — reviews, theses, essays, course material, internship reports and coursework. Filter by type, or explore by topic.
-
Cours - Du Machine Learning aux réseaux convolutifs
A ten-hour machine learning course for third-year bachelor students: from the supervised learning pipeline through to CNNs, U-Net and variational auto-encoders.
-
A Moment to Talk About AI
Why frontier AI labs need binding rules, and why the generation that understands these systems must help write them — a follow-up to the WAIRO proposal, written against the events of September 2026.
-
PhD Review - Time–Frequency Analysis Methods for Feature Engineering
A review of time–frequency methods for feature engineering on non-stationary vibration signals.
Inria -
PhD Review - Time Domain Methods for Fatigue and Damage Estimation
A review of time-domain fatigue estimation, from cycle counting to cumulative damage models.
Inria -
PhD Review - Frequency Domain Methods for Fatigue and Damage Estimation
A review of frequency-domain methods for fatigue and damage estimation, and the assumptions each one makes.
Inria -
WAIRO - A Global AI Regulation Framework
A proposal for a World Artificial Intelligence Regulation Office: an independent, binding global authority for AI oversight, with a departmental structure, a funding model and a phased path to adoption.
-
Apprenticeship Thesis - Leveraging Machine Learning for Dynamic and Intelligent Cloud Commitment Recommendations
Apprenticeship thesis: machine learning for dynamic and intelligent cloud commitment recommendations, at Sudo Group.
Sudo Group -
Presentation - Work and Project Overview
The work and project overview presented for the Inria PhD application.
Université de Lille -
Thesis - Load-balancing and Task Allocation in Dynamic Multi-Agent Systems
Master's thesis: load-balancing and task allocation in dynamic multi-agent systems. Supervised by Pr. Maxime Morge.
Université de Lille -
Natural Language Processing - Transformers Workshop 1
Transformers workshop: attention from first principles, then fine-tuning in practice.
Université de Lille -
Multi-Agent Systems
Multi-agent systems: emergence, coordination and what happens when agents share an environment.
Université de Lille -
BlackRock - Geopolitical Influence of a P.E. conglomerate.
A short report on BlackRock's geopolitical influence, sourced from verifiable reporting.
-
Deep Learning - Denoising Autoencoder
Building a denoising autoencoder, and reading the latent space it learns.
Université de Lille -
Dimension Reduction - Uniform Manifold Approximation and Projection
Uniform Manifold Approximation and Projection: what UMAP preserves, and what it does not.
Université de Lille -
Cybersecurity3 - Introduction to secure systems - University of Lille
Introduction to secure systems, part 3: putting the attack chain together.
Université de Lille -
Cybersecurity2 - Introduction to secure systems - University of Lille
Introduction to secure systems, part 2: buffer overflows and defensive compilation.
Université de Lille -
Cybersecurity1 - Introduction to secure systems - University of Lille
Introduction to secure systems, part 1: memory, privilege and the shape of an exploit.
Université de Lille -
Data Science - Facial Landmark Emotion Classification - University of Lille
Classifying emotion from facial landmarks — feature engineering on geometry rather than pixels.
Université de Lille -
Project XSA (X Sentiment Analysis) - University of Lille
Project XSA: sentiment analysis on X, from data collection through to a deployed classifier.
Université de Lille -
IIR - Reinforcement learning and how general-purpose learning and communication methods can solve complex environments
How general-purpose learning and communication methods let reinforcement learning agents solve complex environments.
Université de Lille -
Algorithms and Complexity - University of Lille
Algorithms and complexity: designing algorithms and proving what they cost.
Université de Lille -
Internship - Objective Evaluation on Benchmarked Urban Environment LiDAR Data
Objective evaluation of a 3D point cloud classification algorithm on benchmarked urban LiDAR data, at CNRS OSUR.
-
Internship - Semantic Segmentation State Of The Art For Ultra High Resolution UAV Imagery
Semantic segmentation for ultra-high-resolution UAV imagery: state of the art and experimental comparison, at CNRS LETG.
-
Differential Calculus - Dynamical Systems and Stability
Dynamical systems and stability analysis, illustrated on an SEIR epidemic model.
UTC -
Differential Calculus - Solving Systems of Algebraic Equations
Solving systems of algebraic equations, applied to the inverse kinematics of a robotic arm.
UTC -
Differential Calculus - Piecewise Linear Regression
Piecewise linear regression: fitting models with breakpoints and choosing where they go.
-
Dynamic System Simulation - Car lane insertion
Simulating car lane insertion as a dynamic system, in MATLAB.
UTC -
Differential Calculus - Splines, Runge Phenomenon
Splines and the Runge phenomenon — why high-degree polynomial interpolation fails.
UTC -
Project - State of the Art on Intelligence to manipulate Covid-19 Data.
State of the art on the methods used to exploit Covid-19 data, with a Dataiku pipeline built on top.
Université Bretagne Sud -
Linear Regression and Simulation Methods Final Exam
Final exam on linear regression and simulation methods, worked through end to end.
Université Bretagne Sud -
Python and Machine Learning - Deep Learning Introduction
Introduction to deep learning: from the perceptron to a trained network, in Python.
Université Bretagne Sud -
Hands On - Clustering data II
Second clustering session: hierarchical methods and validation of the number of clusters.
Université Bretagne Sud -
Statistical Analysis of Cardiovascular Closure Systems related to health state 1 month after operation - EOT Assignement
Statistical analysis of cardiovascular closure systems against patient health one month after operation.
Université Bretagne Sud -
Personal Project - Reinforcement Learning applied to robotic agent (Bipedal Walker)
Teaching a Bipedal Walker to walk — a self-directed reinforcement learning project on a continuous control task.
-
Hands On - Clustering data
Hands-on clustering: choice of distance, partitioning methods, and reading the output.
Université Bretagne Sud -
Tamagotchi in Java - EOT Assignment
A five-person Java project: object design, state machines and a playable Tamagotchi.
Université Bretagne Sud -
In depth analysis of medical data to unveil heart disease - EOT Assignement
In-depth statistical analysis of medical records to surface the factors associated with heart disease.
Université Bretagne Sud -
Bank loan payment default prediction - EOT Assignement
Predicting bank loan payment default in R, from exploratory analysis through to model comparison.
Université Bretagne Sud -
Internship - Building a predictive model for road surface condition, Accenture Technology.
First engineering placement: a predictive model for road surface condition, built on an industrial Dataiku pipeline.
Accenture
Nothing under that filter.