Science & AI
What I'm studying, what I've built, and what I'm still working out.
Projects
neural-network-numbers
A neural network built from scratch that reads hand drawn digits, wired to a 3D visualisation. You draw a number and watch the activations move through the layers.
air-harp
An instrument you play in the air. MediaPipe tracks your fingertip through the webcam, virtual strings hang in the frame, and crossing one plucks it. Multi channel audio, so notes overlap properly.
ML_library
A machine learning and matrix library in pure C. Matrix operations, ReLU and sigmoid, dense layers, forward propagation and MSE loss, with nothing underneath it but the language.
vector-search-engine
A semantic search engine written from the linear algebra up. 384 dimensional BERT embeddings, normalised so that retrieval is one matrix multiply, behind a FastAPI service.
omni-analyst
An agentic analyst built on the Mistral SDK. It takes a question, searches the live web, writes and runs its own Python to chart what it finds, then writes the report.
Research
De la propagation avant à la rétropropagation : une analyse des mécanismes d'apprentissage dans les réseaux de neurones multicouches
S. Sutton · Aix-Marseille Université · HAL, 2026
A walk through the mathematics of learning in multilayer networks, from Rosenblatt's perceptron to backpropagation. Written in French; my first published paper.
Education
- 2026—2027Paris · FR
Université Paris-Saclay
M1 — Data, Knowledge & Hybrid AI (DKAI)
A two year programme combining data science and artificial intelligence, taught entirely in English. It covers machine learning and big data alongside the different paradigms of AI, from symbolic and agent-based through to hybrid and generative, with projects, challenges and internships in research labs or industry.
CourseworkFoundations of knowledge graphs, mathematics for data science, machine learning, deep learning, large-scale data management and data quality, distributed query processing, constraint programming, algorithmic foundations of data mining, agent-based systems, interactive information visualisation, trustworthy AI.
- 2026—2027Marseille · FR
Aix-Marseille Université
Bachelor of Science — Mathematics
CourseworkEuclidean spaces & geometry, group theory, multivariable calculus & differential equations, topology, Fourier & power series, Markov chains, inferential statistics, numerical analysis.
- 2023—2026Marseille · FR
Aix-Marseille Université
Bachelor of Science — Mathematics & Computer Science
I went back to the fundamentals because I wanted to know how a computer actually works. I stayed for the whole degree.
CourseworkMathematical logic, real analysis (sequences, series, Riemann integration), linear algebra & endomorphism reduction, probability theory, algorithms & data structures, automata & formal languages, databases, compilation, computability theory, operating systems, Bash scripting, natural language processing.
- 2021—2023London · UK
Middlesex University
BA (Hons) Music Production & Sound Engineering
A two year accelerated degree at Point Blank, validated by Middlesex.
What I'm working toward
- 01
Research
I want to work on AI itself, not only use the tools other people build. That means going deep enough to add something rather than assembling what already exists.
- 02
Applied to something real
Medicine, music, or a field I haven't run into yet. I genuinely don't know which one yet and I'm still figuring it out. I would rather say that than pretend I have had a plan since I was twelve.
- 03
Curiosity, mostly
This is the real reason for the pile of degrees. I get interested in something, start taking it apart, and end up studying it properly. Sound engineering went that way, then maths, then AI. I expect the next one will too.
No black boxes
If I use something I want to know why it works, and I want to be able to show you. That is most of what I am doing here.