Robin Leman

About

Research Engineer in ML and Robotics. MSc in Computational Statistics and Machine Learning at UCL, researched causal ML at the Gatsby Computational Neuroscience Unit. Previously: 3+ years shipping AAA games at Electronic Arts, Relic Entertainment (SEGA), Microsoft Xbox, and Ubisoft.

Research

• Doubly Robust Deep Kernel Tests for Structured Outcomes. MSc thesis at the Gatsby Unit, supervised by Houssam Zenati and Arthur Gretton.

Projects

• 

  Cosal

  A platform for personal AI agents where each agent runs in its own Firecracker microVM with persistent state, a scheduler, and a tool runtime.
  Rust, TypeScript, Next.js, Firecracker
  [Website] [Code]

• 

  Autocurricula

  A pip-installable LLM-driven interview practice platform with adaptive problem generation, sandboxed code execution, and per-problem scaffolding.
  Python, FastAPI, Claude
  [PyPI] [Code]

• 

  Pico-Banana

  A PyTorch implementation of a Denoising Diffusion Probabilistic Model (DDPM) trained on FashionMNIST.
  Python, PyTorch
  [Code]

• 

  Cart-Pole RL Agent

  A reinforcement learning agent built from scratch in C++, trained on the cart-pole environment.
  C++
  [Code]

• 

  Tiny-LLM

  A lightweight transformer-based language model implementation in PyTorch.
  Python, PyTorch
  [Code]

• 

  CUDA-Accelerated Raytracer

  A high-performance GPU-accelerated raytracer built with CUDA and OpenGL.
  C++, CUDA, OpenGL
  [Code]

• 

  Debris Disk Renderer

  A C++ 3D rendering engine for visualizing and simulating debris disks around solar systems.
  C++, OpenGL
  [Code]