Projects
Partinet
Cryo-EM particle-picking with AICLI and YOLOv10 integration for PyTorch cryo-EM particle-picking tool developed with researchers.
XGBoost pipeline parallelisation
Nextflow-based HPC hyperparameter tuning pipelineScaled XGBoost pipeline on HPC and added GPU support via RapidsAI (6x faster). Hosted on Seqera Platform.
AI-powered backyard bird watcher
YOLO + OpenCV bird species trackerNPU inference with C++ pipeline to track bird species in backyard, with semi-auto data collection and training.
High-performance improvements to in-house SPH code
Accelerating an in-house Fortran SPH codeMPI parallelisation (3,250x speedup on 4,096 cores) and algorithmic/vectorization upgrades for a further ~8x speedup. Enabled GPU support via C++/Kokkos/Thrust.