High-performance computing postgraduate personal statement example

Personal statement example

My interest in high-performance computing began with a piece of code that worked perfectly and was uselessly slow. In my second year I wrote a two-dimensional heat diffusion solver in C for a numerical methods module. It produced sensible results on a small grid, but when I refined the mesh to see the boundary behaviour more clearly, a single run took most of an afternoon. My first instinct was to look for a cleverer algorithm. What actually helped was reading about memory layout and discovering that my nested loops were traversing the array in the wrong order, so almost every access missed the cache. Reordering two lines of code made the program several times faster. That gap between what I had been taught about complexity and what the machine actually rewarded is what I want to spend the next stage of my study on.

My final-year project built on that. I extended a fluid-flow simulation to run in parallel, first with OpenMP across the cores of a single node and then with MPI across four nodes of my department's teaching cluster. The shared-memory version was straightforward; the distributed version taught me far more. I spent weeks on halo exchange, and longer still on understanding why my strong scaling curve flattened at sixteen processes. Profiling showed that communication was not the main problem: load imbalance was, because my domain decomposition split the grid evenly by cells rather than by work, and the regions near the obstacle required more iterations to converge. Rebalancing the decomposition recovered a reasonable portion of the loss. I also learned the discipline of measuring properly, running each configuration several times and reporting variation rather than best-case figures, after my supervisor pointed out that my early numbers were suspiciously tidy. The project was marked well, and I was pleased that the write-up was judged mainly on the honesty of the analysis rather than the headline speed-up.

Alongside my degree I have worked weekends on the service desk of a garden centre, dealing with deliveries, refunds and stock queries. It is not technical work, but it has made me comfortable explaining something unwelcome clearly and calmly, and it has taught me to keep a queue moving without cutting corners. I have also helped run a weekly coding club at my local library for children aged around nine to thirteen. Preparing a session on sorting for that group forced me to strip an idea back to its essentials, and I now find that explaining a parallel algorithm to a peer follows a similar pattern: establish what each part is responsible for before worrying about the detail.

My strongest technical foundations are in C, Python and Linux, with coursework in operating systems, computer architecture and numerical computation. I am aware of the gaps. My linear algebra is adequate rather than fluent, and I have only read about GPU programming rather than written anything substantial in CUDA. I have been working through Hager and Wellein's Introduction to High Performance Computing for Scientists and Engineers in my own time, which has given me a much clearer mental model of bandwidth limits and why some of my earlier optimisation attempts were bound to fail. A taught postgraduate course would give me structured time on accelerators, parallel numerical methods and the tools for performance analysis, with access to hardware I cannot replicate at home.

In the longer term I would like to work as part of a team supporting research computing, helping scientists get more from the machines they already have. That work seems to sit exactly where I am happiest: close to the hardware, and in conversation with people whose problems are not primarily about computing at all.

Review this personal statement

Latest reviews