I’ve always liked working across hardware and software. These projects span systematic trading, numerical computing at university, and an early hardware build.
During university
I built and ran a fully systematic cryptocurrency statistical arbitrage strategy. I developed and operated the trading software, built my own physical server, and hosted my own local Git infrastructure.
It was an independently operated, self-hosted system. I owned the work end to end, from the machine it ran on to the code that kept it running. I shut it down before starting my full-time quant role.
During university
I simulated heat diffusion across a square plate using forward and backward Euler methods. CPU and CUDA implementations explored how grid resolution and time step affect stability, accuracy and runtime when estimating the time for the plate’s centre to reach a target temperature.
During university
I investigated how sparse matrix storage formats affect GPU computation, implementing conversion from CSR to ELLPACK-R and developing CPU and CUDA multiplication routines. I benchmarked performance and numerical accuracy on Poisson matrices and Matrix Market examples.
During university
I solved the two-dimensional Poisson equation using sparse finite differences and iterative Jacobi methods. I compared CPU and CUDA implementations against an analytical solution to examine accuracy and convergence as the computational grid was refined.
During university
I studied dense matrix multiplication using Python, NumPy and Numba. Benchmarks explored how just-in-time compilation, CPU parallelisation, memory layout and cache blocking affect performance across different matrix sizes.
Age 15
I assembled the electronics and wrote the Arduino/C code to build a working oscilloscope. It was an early chance to bring hardware and software together in something I could use.