Senior Engineering Manager, Kernel and Virt

DigitalOcean•Seattle, WA

15d•Hybrid

About The Position

Dive in and do the best work of your career at DigitalOcean. Journey alongside a strong community of top talent who are relentless in their drive to build the simplest scalable cloud. If you have a growth mindset, naturally like to think big and bold, and are energized by the fast-paced environment of a true industry disruptor, you’ll find your place here. We value winning together—while learning, having fun, and making a profound difference for the dreamers and builders in the world. We are seeking a Senior Engineering Manager to lead our Inference Orchestration team, driving the strategy, execution, and scaling of our Kubernetes-based AI infrastructure. You will be responsible for balancing business needs with technical excellence, ensuring high throughput, optimal GPU utilization, and robust fault tolerance for our next-generation disaggregated inference, fine-tuning, and training workloads.

Requirements

Proven track record of managing and growing high-performing engineering teams, preferably within a distributed systems or infrastructure domain.
Deep expertise in Kubernetes at scale and a strong foundational understanding of the core challenges in AI workload orchestration, scheduling, and resource management.
Strategic knowledge of GPU architectures (NVIDIA and/or AMD), interconnects (like NVLink), and hardware topology and their direct impact on AI training and inference performance.
Experience in balancing performance against cost, applying principles like Dominant Resource Fairness (DRF), and directing strategies for maximizing cluster efficiency.
Familiarity with concepts in container runtime internals, system isolation, and security contexts to manage risk in shared infrastructure.
Strong understanding of modern LLM serving architectures, disaggregation patterns, and common serving engines (e.g., vLLM, Triton, SGLang).
Expertise in defining, tracking, and operationalizing deep infrastructure and inference metrics (e.g., TTFT, TPOT) to drive performance improvements and meet service level objectives.

Responsibilities

Recruit, mentor, and coach engineers on the team, fostering a culture of ownership, technical excellence, and continuous improvement.
Own the team's project execution, translating high-level business goals into clear technical roadmaps, measurable milestones, and successful, on-time delivery.
Collaborate with Product Management, other engineering teams, and key stakeholders to align priorities, manage dependencies, and communicate progress and risks.
Ensure the production health, stability, and on-call rotation of all services owned by the Inference Orchestration team.
Define the technical roadmap and oversee the architecture of high-throughput scheduling systems for massive Kubernetes clusters (1,000+ nodes, 10,000+ pods), focusing on scalability techniques like multi-scheduler architectures and batch dispatching.
Eliminate GPU waste in multi-tenant environments by implementing fractional GPU allocation, leveraging mechanisms like KAI-Scheduler's Reservation Pods or hard-isolation tools like HAMi, and configuring time-based fairshare scheduling to balance over-quota pool access.
Implement and manage disaggregated AI inference pipelines using frameworks like NVIDIA Grove, coordinating multicomponent deployments (e.g., prefill leaders, decode workers, KV routers) with multilevel autoscaling and explicit startup ordering.
Deploy topology-aware scheduling to align pod placement with physical hardware dimensions, such as NVLink connections, PCIe lanes, and NUMA nodes, minimizing communication latency for multi-GPU operations.
Drive initiatives to enhance overall cluster performance, including optimizing scheduling latency, API server load, and implementing fault tolerance mechanisms like Checkpoint/Restore for long-running AI training jobs.
Orchestrate efficient model weight distribution using OCI Image Volumes and implement Checkpoint/Restore capabilities (via CRIU and NVIDIA cuda-checkpoint) for long-running training fault recovery.
Define and enforce security best practices for AI workloads, ensuring multi-layered isolation environments and agent sandboxes are deployed to safely execute untrusted code (e.g., using Kata Containers, gVisor, or microVMs).