Engineering Manager, Private Cloud Compute Inference Routing at Apple

Job description

Routing LLM inference at Apple's scale takes systems that adapt to changing demand, available capacity and failures while keeping latency low and accelerator utilization high. As the fleet grows across regions and cells, routing must account for capacity and health across the whole fleet, not just a single site. PCC adds a basic architectural requirement: request content stays encrypted outside trusted compute nodes. You'll lead the design of routing and scheduling strategies within this privacy model. Together with inference and security teams, you'll evaluate techniques such as cache-aware request placement and prefill/decode disaggregation. Your scope includes the inference AI gateway layer, where requests are authenticated, metered, prioritized and streamed back to clients, all without exposing request content. This work has broad reach. Every improvement to routing makes Siri AI and Apple Intelligence faster and more reliable, and makes better use of Apple's inference capacity. In your first year, you'll be expected to grow a high-performing organization, set a clear technical roadmap, and measurably improve routing latency, availability and fleet utilization. Extensive software engineering experience with a demonstrated track record of leading teams that design, build, and operate large-scale distributed systems in production environments—whether through direct management, indirect oversight, or a combination thereof Proven ability to recruit, develop, and retain high-performing engineers and leaders Experience delivering API gateways (preferably inference AI gateways), request routing, load balancing, scheduling, or traffic management systems at scale Expertise of LLM inference serving, including request routing, KV-cache placement and reuse, prefill/decode disaggregation, and the tradeoffs between latency, throughput and accelerator use A record of shipping highly reliable systems in high-availability environments, including leading incident management, on-call and reliability practices Strong technical depth: capable of contributing meaningfully to architectural and design discussions with senior engineers, with a server development background in Java, C/C++, Go or Swift Ability to drive alignment and execution across teams in complex, matrixed environments Bachelor’s degree in an engineering discipline or equivalent professional experience Proven leadership in developing LLM inference serving or routing platforms (e.g., llm-d, NVIDIA Dynamo, vLLM, SGLang, or managed inference services) Experience building or running inference AI gateways and their core behaviors: token-based rate limiting and usage metering, multi-model routing with fallback, retries and circuit breaking, and streamingresponse handling Track record of scaling teams and organizations during rapid growth while maintaining top-tier hiring and engineering standards Experience delivering capacity tiers, reservations or QoS for high-volume workloads, and improving accelerator utilization through cross-region, capacity-aware routing Understanding of privacy-preserving system design: TLS, X.509, key management, Oblivious HTTP, remote attestation Experience designing and maturing observability practices, including SLO/SLI frameworks, distributed tracing and on-call culture Experience with container orchestration systems such as Kubernetes Demonstrated ability to build psychologically safe, high-trust team environments that attract and retain top talent Experience formulating multi-year roadmaps in collaboration with product, research, and senior leadership within large organizations that prioritize quality and reliability

Requirements

Must-haves

Proven ability to recruit, develop, and retain high-performing engineers and leaders
Experience delivering API gateways (preferably inference AI gateways), request routing, load balancing, scheduling, or traffic management systems at scale
Expertise of LLM inference serving, including request routing, KV-cache placement and reuse, prefill/decode disaggregation, and the tradeoffs between latency, throughput and accelerator use
A record of shipping highly reliable systems in high-availability environments, including leading incident management, on-call and reliability practices
Strong technical depth: capable of contributing meaningfully to architectural and design discussions with senior engineers, with a server development background in Java, C/C++, Go or Swift
Ability to drive alignment and execution across teams in complex, matrixed environments
Bachelor’s degree in an engineering discipline or equivalent professional experience
Proven leadership in developing LLM inference serving or routing platforms (e.g., llm-d, NVIDIA Dynamo, vLLM, SGLang, or managed inference services)
Experience building or running inference AI gateways and their core behaviors: token-based rate limiting and usage metering, multi-model routing with fallback, retries and circuit breaking, and streamingresponse handling
Track record of scaling teams and organizations during rapid growth while maintaining top-tier hiring and engineering standards
Experience delivering capacity tiers, reservations or QoS for high-volume workloads, and improving accelerator utilization through cross-region, capacity-aware routing
Understanding of privacy-preserving system design: TLS, X.509, key management, Oblivious HTTP, remote attestation
Experience designing and maturing observability practices, including SLO/SLI frameworks, distributed tracing and on-call culture
Experience with container orchestration systems such as Kubernetes
Demonstrated ability to build psychologically safe, high-trust team environments that attract and retain top talent
Experience formulating multi-year roadmaps in collaboration with product, research, and senior leadership within large organizations that prioritize quality and reliability

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

llmaiapijavagoswiftkubernetes

Platform: apple
Effort: medium