李乐岩 Email: leyanpower@gmail.com / 3073751449@qq.com Phone: +86 18805506898 GitHub: https://github.com/lionleepower --- University of Edinburgh（爱丁堡大学） 高性能计算与数据科学 硕士（MSc） 2025年09月 – 至今（预计2026年09月毕业） ________________________________________ University of Liverpool（利物浦大学） 计算机科学 学士（BSc，一级荣誉学位） 2023年09月 – 2025年06月 ________________________________________ 西交利物浦大学（Xi’an Jiaotong-Liverpool University） 信息与计算科学 学士 2021年09月 – 2023年06月 --- Summary 高性能计算与数据科学硕士在读，具备扎实的并行计算与分布式系统背景。熟练掌握 C/C++、Python、MPI 与 OpenMP，具有在 ARCHER2 超级计算机上进行大规模实验的实践经验。研究兴趣包括可扩展系统、性能优化，以及高性能计算与机器学习的交叉领域。 --- Technical Skills： C/C++、Python、SQL ; MPI、OpenMP、PETSc、Slurm、ARCHER2; Linux、Git、VS Code, linaro map ----- Projects 1. PETSc Benchmark and Performance Analysis Suite GitHub: https://github.com/lionleepower/petsc-benchmark-suite Developed a reproducible benchmarking framework for analysing PETSc linear solver performance under different parallel configurations. Designed automated Slurm job pipelines to conduct strong scaling and hybrid MPI+OpenMP experiments on the ARCHER2 supercomputer. Built Python-based analysis tools to process runtime logs, generate CSV datasets, and produce speedup and parallel efficiency visualisations. Performed performance analysis comparing MPI-only and hybrid execution models, identifying communication overhead and scalability limits. 2. HPC Mini Applications (MPI / OpenMP) — in progress Implemented a collection of classical high-performance computing mini-applications in C, including: • 2D heat diffusion simulation using stencil computation • parallel matrix multiplication • N-body simulation Designed MPI-based domain decomposition and halo exchange mechanisms for distributed-memory parallelism. Integrated OpenMP thread-level parallelism and evaluated hybrid MPI+OpenMP configurations. Conducted scaling experiments to analyse parallel efficiency, load balance, and memory behaviour under different process and thread configurations. Value Function Factorisation in Multi-Agent Actor-Critic Methods Investigated factorised critic architectures (VDN, QMIX) within Multi-Agent Proximal Policy Optimization (MAPPO) to improve coordination in cooperative multi-agent environments. Implemented custom MARL algorithms using Python and PyTorch based on the PyMARL / ePyMARL framework. Conducted experimental evaluation in Matrix Game and Predator–Prey environments, analysing convergence behaviour and coordination performance. Explored improvements to reward shaping and critic factorisation strategies to enhance learning stability. ----------------------------------------------------------------------------------------------------------- Research Experience String Art Generator Based on Radon Transform Summer Undergraduate Research Programme Jun 2024 – Aug 2024 Developed an algorithmic pipeline that converts images into string-based artwork using Radon transform techniques. Implemented image preprocessing and contrast enhancement algorithms to improve feature extraction. Designed a parameter control interface enabling interactive adjustment of artistic generation parameters. ------------------------------------------------------------------------------------------- Publication Li, Leyan (2024) Convolutional Neural Networks (CNNs)-based Medical Image Analysis International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2024)