Heterogeneous AI Interconnects Engineer

United States Digital Space LLC; Greater London; On-site; GBP 101,000 - 192,000; Full time

Benefits offered by this job

Equity & Ownership

Private healthcare

Visa sponsorship

Relocation benefits

Job summary

Intelligent Systems Company in London is building a cutting-edge communication system for heterogeneous accelerators. You will design and implement the networking and data movement layers to enable efficient cross-vendor cooperation, addressing latency and bottlenecks across the hardware and software stack.

The role requires deep expertise in networking, RDMA/Ethernet fabrics, and performance tuning, with a bias toward simplifying complex architectures to meet workload requirements.

Qualifications

• Deep understanding of networking and communication systems across protocols, hardware, memory and software layers.
• Experience with high-performance data movement such as RDMA, Ethernet fabrics, accelerator interconnects, programmable networking.
• Strong systems performance instincts, ability to trace bottlenecks across abstraction layers.
• Demonstrable tendency to question abstractions and simplify systems for workload requirements.

Responsibilities

• Characterise and simplify communication paths between accelerators, tracing latency and overhead.
• Develop faster data movement across heterogeneous hardware (RDMA, DPUs, NICs).
• Build mechanisms for transferring model state across accelerator types and execution strategies.
• Evaluate scale-up/scale-out fabrics and inform cluster architecture for heterogeneous communication.

Skills

Networking & IPC

RDMA & Ethernet

Systems performance

Abstraction & generalisation

Job description

Intelligent Systems Company in London is building a cutting-edge communication system for heterogeneous accelerators. You will design and implement the networking and data movement layers to enable efficient cross-vendor cooperation, addressing latency and bottlenecks across the hardware and software stack.

The role requires deep expertise in networking, RDMA/Ethernet fabrics, and performance tuning, with a bias toward simplifying complex architectures to meet workload requirements.