Generating images from a description by inverting a vision-language model (Qwen3.5). The image is optimized until the model describes it with exactly the given sentence. The multigrid parameterization from ODIL gives recognizable objects instead of adversarial noise. README with examples.
ODIL (Optimizing a DIscrete Loss) is a method and Python framework for solving inverse problems for partial differential equations, which is orders of magnitude faster than PINN (physics-informed neural networks). Article about the method.
Distributed multiphysics solver in C++ with MPI for simulating multiphase flow with bubbles and electrochemical reactors. The solver performed the largest simulations of foaming by breakup and mixing of air in water. Article about foam simulations.
Automatic differentiation framework in C++ with GPU support through OpenCL.
Visual materials for a class on numerical methods, lectured in 2022.
Game with particles and portals in C++.
Prototype operating system in x86 assembly for a school competition in 2008.