ChinaHPC Engineer
2 мес. назад
ChinaWorldwide
gpuhigh performance computingcluster managementdistributed computingmonitoring
HPC Engineer role responsible for designing, operating, and optimizing large-scale GPU compute clusters for distributed model training and high-performance workloads in a quantitative trading firm.
О компании
- AlphaGrep 是一家全球领先的量化交易公司，专注于股票、商品、外汇及固定收益等资产的算法交易。我们在国际市场拥有显著份额，依托自主开发的超低延迟系统与严格的风控体系，持续构建高效能策略。
- AlphaGrep is a leading global quantitative trading firm specializing in algorithmic strategies across equities, commodities, FX, and fixed income. We hold significant market share internationally, powered by proprietary low-latency infrastructure and robust risk controls.
- AlphaGrep China 是专注于中国市场人民币资产管理的机构，服务对象涵盖机构投资人、家族办公室与高净值客户。我们深耕中国资本市场，涵盖股票及衍生品等多类资产，结合全球量化研究体系与本地实战经验，构建多元交易策略，致力于实现长期稳健增长。
- AlphaGrep China is a dedicated RMB asset management platform focused on the Chinese market, serving institutional investors, family offices, and high-net-worth individuals. We leverage deep expertise in China’s equity and derivatives markets, together with AlphaGrep’s global quantitative research capabilities, to deliver diversified strategies targeting long-term and stable returns.
-
Обязанности
- 我们正在寻找一位 HPC / GPU 集群工程师，协助设计、运营并持续优化公司用于分布式模型训练及高性能计算的大规模 GPU 计算环境。您将端到端负责集群的性能与稳定性——从 GPU 硬件与互联网络，到存储层、调度系统、监控平台，以及保障数百块加速器高效运行的全套工具链。
- We are looking for an HPC / GPU Cluster Engineer to help design, operate, and continuously optimize our large-scale GPU compute environment used for distributed model training and high-performance workloads. You will own the performance and reliability of the cluster end to end — from the GPUs and interconnect fabric up through the storage layer, scheduler, monitoring, and the tooling that keeps hundreds of accelerators running efficiently.
-
Требования
- 具备生产环境下大规模 GPU 或 HPC 集群的运营经验，能够系统性地识别并解决性能瓶颈。
- Experience operating large GPU or HPC clusters in production, identifying performance bottlenecks and resolving them systematically.
- 具备 RDMA 编程及底层调试的实战经验，包括 RDMA verbs / libibverbs、内存注册、队列对、完成队列、RDMA 读写、发送/接收操作及性能基准测试。
- Hands-on experience with RDMA programming and low-level RDMA debugging, including RDMA verbs / libibverbs, memory registration, queue pairs, completion queues, RDMA read/write, send/recv, and performance benchmarking.
- 熟悉分布式训练的核心底层技术，包括 InfiniBand 和/或 RoCEv2、GPUDirect RDMA、GPUDirect Storage、NCCL、CUDA 驱动及 OFED。
- Practical experience with the technologies underpinning distributed training, including InfiniBand and/or RoCEv2, GPUDirect RDMA, GPUDirect Storage, NCCL, CUDA drivers, and OFED.
- 具备 Slurm 等调度系统的使用经验，包括安装配置、队列/分区管理、作业故障排查及监控。
- Experience with workload schedulers such as Slurm, including setup, configuration, queue/partition management, job troubleshooting, and monitoring.
- 具备高性能共享存储或并行/分布式文件系统的搭建与管理经验，如 Lustre、BeeGFS、WEKA、VAST、DDN/ExaScaler 等。
- Experience setting up and managing high-performance shared storage or parallel/distributed filesystems such as Lustre, BeeGFS, WEKA, VAST, DDN/ExaScaler, or similar systems.
- 熟练掌握 Python、Bash，优先具备 C/C++ 能力，用于集群自动化、诊断、基准测试及监控开发。
- Solid scripting/programming ability in Python, Bash, and preferably C/C++, for cluster automation, diagnostics, benchmarking, and monitoring.
- 熟悉 HPC 网络设计，包括阻塞与非阻塞网络架构、InfiniBand、高性能以太网、网络拓扑、拥塞控制及端到端带宽/延迟故障排查。
- Familiarity with HPC networking, including blocking vs. non-blocking fabric design, InfiniBand, high-performance Ethernet, topology, congestion, and end-to-end bandwidth/latency troubleshooting.
- 扎实的 Linux 系统管理能力，包括网络配置、文件系统管理、内核/驱动问题处理、进程与资源管理及性能调试。Strong Linux system administration skills, including networking, filesystems, kernel/driver issues, process/resource management, and performance debugging.
-
Другое
- 信任是团队协作的根基
- 我们鼓励坦诚沟通与主动承担，让每一位成员都能在安全感中成长，自主决策、共同前行。这份信任源于彼此支持与并肩作战，是我们最珍贵的团队资产。
- Trust is the foundation of collaboration.
- We foster open communication and proactive ownership, empowering every team member to grow with a strong sense of security, make autonomous decisions, and move forward together. This trust—built through mutual support and shared commitment—is our most valued asset.
- 优秀的团队成员，我们汇聚了工程师、数学家、统计学家，保持好奇心，乐在其中。
- Great People. We’re curious engineers, mathematicians, statisticians and like to have fun while achieving our goals.
- 透明的组织架构，我们重视每一位成员的想法与贡献。
- Transparent Structure. Our employees know that we value their ideas and contributions.
- 轻松的办公环境，无等级文化，常有团建、聚会与休闲活动。
- Relaxed Environment. Flat organization with yearly offsites, happy hours, and more.
- 健康福利支持，健身补贴、零食饮品、充足年假。
- Health & Wellness Programs. Gym membership, stocked kitchen, and generous vacation.