Principal AI Back-End Network Systems Lead

Job d:

Location:

United States, California, Mountain View Salary: USD $142,800 - $274,800 per year

Employment type:

Full-Time Work site: 3 days / week in-office

Role type:
Individual Contributor Travel:
Less than 25%
Profession:
Hardware Engineering Discipline:
Hardware Engineering Company:
Microsoft Overview Microsoft  Silicon, Cloud Hardware, and Infrastructure Engineering (SCHIE) is the team behind Microsoft's expanding Cloud Infrastructure and is responsible for powering Microsoft's Intelligent Cloud mission. SCHIE delivers the core infrastructure and foundational technologies that support Microsoft Azure and other online services globally. Our focus is on smart growth, high efficiency, reliability, and a trusted experience for customers and partners worldwide.

The Platform Systems Engineering (PSE) team is seeking a Principal AI Back-End Network Systems Lead to serve as the technical lead and end-to-end product development owner for Back End Network across both scale-up and scale-out. The role will lead the product from requirements and architecture through design, integration, validation, manufacturing readiness, deployment, and fleet learning - spanning silicon, network adapters, switches, interconnects, optics/electrical channels, and the fabric as a complete system.

The successful candidate will bring deep expertise in one or more AI/ML scale-up or scale-out networking domains and the product-level thinking to connect decisions across the full system. This lead will work from physical and data-link technologies through network and transport behavior to collectives, workloads, modeling, observability, and datacenter deployment, creating clarity across architecture, execution, quality, performance, reliability, power, cost, and schedule.

#azure #MAIA #AI/ML #Networking #Scale Up #Scale Out Responsibilities End-to-End Product Ownership & Architecture Own the Back End Network product definition and execution plan across scale-up and scale-out, from customer/workload needs and product requirements through architecture, development, qualification, deployment, and fleet learning. Translate AI training and inference requirements into measurable system targets for bandwidth, latency, scale, collective efficiency, reliability, availability, serviceability, power, thermal behavior, cost, security, and schedule.

Make and drive cross-system architecture decisions, interface contracts, feature priorities, and risk tradeoffs; maintain clear ownership, milestones, entry/exit criteria, and product readiness metrics. Lead NUDD (new, unique, different, and difficult) technology assessments, prototypes, make/buy/partner decisions, and mitigation plans for the highest product risks. Scale-Up/Scale-Out Systems, Silicon, Switch & Fabric Lead architecture and integration across accelerator and networking silicon, NICs/DPUs, PHYs/retimers, switch ASICs, switches, cables/backplanes, optics, and fabric topology.

Define scalable scale-up and scale-out architectures, including topology, radix, reach, link speeds, routing, path diversity, over subscription, failure domains, and resiliency. Drive silicon, board, switch, firmware, network operating system, and fabric requirements so that components operate as one high-performance and serviceable product. Partner with silicon, switch, optical, cable, connector, system, manufacturing, and supplier teams to deliver designs from concept through production and datacenter integration.

Optical/Electrical & Cross-Layer Integration Oversee electrical and optical technologies for high-speed links, including signal integrity, Ser Des/PHY, link training, FEC, reach, power, thermal, mechanical, manufacturability, and interoperability tradeoffs. Drive end-to-end behavior across the relevant OSI layers - physical, data link, network, and transport - and the system/runtime interfaces above them. Define and validate Ethernet/RDMA or other scale-up/scale-out protocol behavior, including flow control, congestion control, QoS, routing, packet loss/recovery, telemetry, and security.

Ensure hardware, firmware, drivers, network operating systems, management software, orchestration, and datacenter operations are aligned to common specifications and concepts of operation. Collectives, Workloads, Modeling & Performance Connect network architecture to distributed AI workloads, parallelism strategies, traffic patterns, and collective operations such as all-reduce, all-gather, reduce-scatter, and all-to-all. Establish…