Nathan Ben Soussan
Concurrency, parallel computing and machine learning
Bio
I'm Nathan, a computer science engineering student at Grenoble INP – Ensimag, a tier-1 engineering school in France, ranked first for computer science by Le Figaro in 2026. I'm interested in what happens when a computer does many things at once: concurrency, parallel computing, and machine learning, where the two meet in one very large, very parallel computation.
I got into computers as a kid because of my older brother, and I've been trying to understand how they work ever since, usually by rebuilding things from scratch until they make sense. Away from the keyboard: chess, sport, and too much tech news.
Internship · 2027
I'm looking for one from mid-May to mid-September: four months, in parallel computing, systems or machine learning. Get in touch
Selected work
- YOCO
An end-to-end machine learning pipeline that reads the FEN notation of a real chess game, from dataset creation to CNN training.
- Tokenizers
BPE, WordPiece and Unigram implemented from scratch, with a guide on how they differ.
- Clické
A macOS menu bar screenshot app with clipboard copy, a built-in editor and export presets.
On your machine
| workers | time | speedup | against perfect scaling |
|---|
Press run and the page starts one Web Worker per logical core, then times the same integer loop split 1, 2, 4… ways. The tick on each bar is perfect scaling; a bar that stops short of it, or runs past it, is your hardware: performance and efficiency cores, clocks that change as cores wake up, whatever else the system is running. The serial fraction is the Karp–Flatt metric those timings imply. Nothing leaves the page.