I’m a computational physicist and software engineer with expertise in GPU programming (CUDA/C++) and high-performance numerical algorithms. I designed and optimized a CUDA-based Fast Multipole Method using kd-trees as part of my MSc thesis research. I have experience with many programming languages and ML frameworks like PyTorch and JAX through remote collaborations, and I have broad theoretical grounding in numerical methods and complex systems.
I currently work as a freelance software engineering consultant. My responsibilities include:
- reproducing, test and analyse real-world bug fixes and features across 20+ large open-source codebases, including SPIRV-Tools and Godot (C++), JAX (Python), OpenGamma Strata (Java, fixed-income pricing) and RxJava (concurrency).
- writing unit and integration tests beyond those included in the original PRs, to verify fixes and edge cases.
- reviewing code produced by AI coding agents for correctness against requirements, safety, codebase conventions and documentation, including agent behaviour in simulated production environments (e.g., avoiding unnecessary service restarts or data mutation).
- configuring and manage Docker environments to reproduce issues and run test suites across C++, Python, Java, JavaScript, TypeScript, Go and Ruby projects.
- I also took part in several pilot projects, e.g., ML engineering benchmark evaluations of AI agent performance on Machine Learning tasks using PyTorch and JAX.
I currently use my GitHub profile to host and maintain my personal and university projects.
My contacts:
My statistics (if they don't load, refresh the page):
