Senior Infrastructure Engineer - GPU Compute

免费注册后解锁"你的匹配分 vs 高分画像"与个人技能缺口。

Boundless is coordinating GPU compute at scale as it becomes a leader in AI. As a Senior Infrastructure Engineer (GPU Compute), you'll build and operate the compute fabric that powers our AI inference workloads — a large, heterogeneous, globally distributed GPU fleet spanning consumer cards (including RTX 5090) and datacenter hardware. Your job is to keep that fleet full, fast, cheap, and always on: orchestrating workloads across regions and providers, squeezing every bit of performance out of the hardware, and driving down cost per GPU-hour. This role rewards engineers who want to go deep on bare-metal and GPU optimization. You should be comfortable operating with a high degree of autonomy, navigating ambiguity, and defaulting to a strong bias for action. What You'll Do GPU Fleet Orchestration: Operate a heterogeneous, multi-region GPU fleet (consumer + datacenter, including RTX 5090) using tools like SkyPilot, Kubernetes/k3s, and cloud + on-prem providers. Build the patterns that let us schedule inference workloads across the entire fleet reliably. Compute Scheduling Utilization: Maximize GPU utilization across inference workloads. Own workload placement across spot, on-prem, and cloud capacity, keeping the "always-on inference substrate" saturated and economical. Bare-Metal GPU Optimization : Go deep on GPU performance — PCIe P2P, ReBAR, NUMA topology (e.g. EPYC SP5), CUDA/driver tuning, memory configuration, and network topology — to push throughput per node. Reliability, Access Observability : Build secure fleet access (Tailscale, Teleport), robust observability and alerting, and zero-downtime rollouts across a distributed node fleet. Cost Optimization: Drive down $/GPU-hr through spot instance management, intelligent workload placement between on-prem and cloud, and resource scheduling — without sacrificing reliability. Requirements 5+ years of infrastructure/DevOps experience operating large-scale production systems Deep expertise in Kubernetes, Docker, and container orchestration at scale Strong Linux systems administration skills Proficiency in infrastructure-as-code tools (Terraform, Ansible, Pulumi) Track record of managing mission-critical, high-throughput systems Strong infrastructure-as-code background in heterogeneous environments Proficiency in at least one common scripting or programming language (Python, Bash, TypeScript, Go, etc.) Comfort navigating ambiguity with a strong bias for action Nice to Have Experience with GPU computing infrastructure (CUDA, bare-metal optimization, kernel tuning) Experience operating ML training or other large-scale distributed compute infrastructure Experience with GPU fleet orchestration (SkyPilot, Ray, Slurm) Familiarity with fleet access and networking tooling (Tailscale, Teleport) Knowledge of network optimization and topology design Experience with multi-region, globally distributed systems Proficiency in Rust or low-level systems programming Experience with on-premises data center operations Additional Requirements Candidates must include a public GitHub profile in their application. The GitHub profile should demonstrate a minimum of 1 year of activity/history. Applications that do not include a GitHub profile, or show insufficient activity, will not be considered. Benefits At Boundless, we take care of our people, because building the future of AI compute starts with an empowered team. Here's what you can expect when you join us: Competitive salary (proposed band b/t US$175k and $250k annually) + equity allocation Health, dental, vision (for U.S. employees; region-adjusted globally) Flexible PTO Professional development and conference travel budget Remote-first with regular off-sites and a high-trust, high-velocity team environment We are a global team, and applicants from around the world are welcome to apply. Originally posted on Himalayas

Boundless正在大规模协调GPU计算，并逐渐成为AI领域的领导者。作为高级基础设施工程师（GPU计算），你将构建并运营为我们的AI推理工作负载提供动力的计算网络——一个大型、异构、全球分布的GPU集群，涵盖消费级显卡（包括RTX 5090）和数据中心硬件。你的工作是让这个集群保持满载、高速、低成本且始终在线：跨区域和跨提供商编排工作负载，从硬件中榨取每一丝性能，并降低每GPU小时的成本。这个角色奖励那些希望深入裸金属和GPU优化的工程师。你应该能够以高度自主性开展工作，驾驭模糊性，并默认具有强烈的行动偏好。你将做什么GPU集群编排：使用SkyPilot、Kubernetes/k3s以及云+本地提供商等工具，运营一个异构、多区域的GPU集群（消费级+数据中心，包括RTX 5090）。构建让我们能够可靠地在整个集群中调度推理工作负载的模式。计算调度利用率：最大化推理工作负载的GPU利用率。负责工作负载在spot、本地和云容量之间的放置，保持“始终在线推理基座”饱和且经济。裸金属GPU优化：深入研究GPU性能——PCIe P2P、ReBAR、NUMA拓扑（例如EPYC SP5）、CUDA/驱动调优、内存配置和网络拓扑——以提升每个节点的吞吐量。可靠性、访问可观测性：构建安全的集群访问（Tailscale、Teleport）、稳健的可观测性和告警，以及跨分布式节点集群的零停机滚动发布。成本优化：通过spot实例管理、在本地和云之间智能放置工作负载以及资源调度来降低$/GPU-hr——同时不牺牲可靠性。要求5年以上基础设施/DevOps经验，运营大规模生产系统深入掌握Kubernetes、Docker和大规模容器编排强大的Linux系统管理技能熟练使用基础设施即代码工具（Terraform、Ansible、Pulumi）有管理任务关键型、高吞吐系统的记录在异构环境中具有强大的基础设施即代码背景精通至少一种常用脚本或编程语言（Python、Bash、TypeScript、Go等）能够从容应对模糊性，并具有强烈的行动偏好加分项有GPU计算基础设施经验（CUDA、裸金属优化、内核调优）有运营ML训练或其他大规模分布式计算基础设施的经验有GPU集群编排经验（SkyPilot、Ray、Slurm）熟悉集群访问和网络工具（Tailscale、Teleport）了解网络优化和拓扑设计有多区域、全球分布式系统经验精通Rust或底层系统编程有本地数据中心运营经验附加要求候选人必须在申请中包含公开的GitHub个人资料。GitHub个人资料应展示至少1年的活动/历史。未包含GitHub个人资料或活动不足的申请将不予考虑。福利在Boundless，我们关爱我们的员工，因为构建AI计算的未来始于一支被赋能的团队。加入我们后，你可以期待：有竞争力的薪资（建议区间为每年175,000美元至250,000美元）+ 股权分配健康、牙科、视力保险（适用于美国员工；全球范围内按地区调整）灵活带薪休假职业发展和会议差旅预算远程优先，定期线下聚会，以及高信任、高速度的团队环境我们是一个全球团队，欢迎世界各地的申请者申请。最初发布于Himalayas

⚑ 岗位信息收录自公开招聘页（Himalayas），以雇主官网为准