Senior High-Performance Engineer - GPU Simulation Cambridge

Dex; Cambridge; On-site; GBP 90,000 - 120,000; Full time

Job summary

Dex is partnering with a Cambridge robotics initiative to build GPU-accelerated simulation and training infrastructure. The role focuses on removing latency in the physics engine, profiling and optimizing both GPU and CPU paths, and shaping the core engine with the founders in a hands-on, iterative process.

We seek a senior IC with deep C++ expertise in latency-sensitive environments such as games, HPC, or physics engines.

Qualifications

• Mastery of modern C++ with focus on low-level profiling, memory layout design, parallelism and deep hardware awareness.
• Experience in environments where latency or throughput is the product (games/HPC/simulation/low-latency systems).
• Proven senior individual contributor capable of raising the technical bar through rigorous code reviews and paired work.

Responsibilities

• Profile and optimize GPU and CPU execution pipelines across the simulation engine, training loops, and edge deployment to reduce latency.
• Architect cache-friendly memory layouts, lock-free structures, and parallel paths tailored to modern hardware limits.
• Test and validate performance gains on physical robots inside the lab.
• Set engineering standards for performance testing and benchmarking via hands-on pairing and code reviews.
• Collaborate with founders to debate architecture and make design calls on the core engine.

Skills

Modern C++ mastery

Low-latency / high-performance systems

Mentoring & code reviews

On-site Cambridge lab

Job description

Dex is partnering with a Cambridge robotics initiative to build GPU-accelerated simulation and training infrastructure. The role focuses on removing latency in the physics engine, profiling and optimizing both GPU and CPU paths, and shaping the core engine with the founders in a hands-on, iterative process.

We seek a senior IC with deep C++ expertise in latency-sensitive environments such as games, HPC, or physics engines.