NLP Performance Engineer

Overview

Join G-Research in London as an NLP Performance Engineer to own large-scale LLM inference performance. You will work with researchers and infrastructure to profile, optimise and deploy efficient NLP workloads on cutting-edge hardware. Your work will shape tools and platforms that accelerate discovery and scale NLP capabilities for the firm. This role combines hands-on engineering with scientific performance optimisation to drive cost-efficient, high-quality inference. You’ll have a clear impact by improving throughput, reliability and scalability of our NLP stack.

Pay / Benefits

• highly competitive compensation
• lunch provided
• 35 days’ annual leave
• 9% company pension contributions
• informal dress code
• comprehensive healthcare and life assurance

Responsibilities

• Profile, benchmark and optimise large-scale LLM inference workloads across compute infrastructure
• Ensure efficient deployment of latest models on diverse GPU architectures and adapt the stack as hardware evolves
• Design and implement inference optimisations while preserving output quality
• Develop reference implementations, libraries and tooling to improve NLP workload efficiency and reliability
• Collaborate with researchers, senior stakeholders and engineers to design optimized solutions
• Work with systems, architecture and platform teams to evolve the compute stack and influence long-term decisions

Key requirements

• Bachelor’s, Master’s or PhD in computer science or equivalent experience
• Proven experience profiling, benchmarking and optimising large-scale LLM inference workloads
• Scientific, evidence-led approach to performance optimisation with reproducible measurement
• Deep understanding of transformer inference (prefill vs decode, KV-cache, attention variants, bottlenecks)
• Hands-on experience with LLM serving frameworks (vLLM, SGLang, TensorRT-LLM, TGI) and PyTorch ecosystem
• Experience with inference optimisation techniques (quantisation, speculative decoding, model parallelism) across modern GPUs
• Strong software engineering skills (Python, CUDA) and building reliable ML systems
• Strong communication skills for cross-team collaboration

• collaborative mindset
• clear communication
• problem-solving with scientific rigor
• vLLM
• SGLang
• TensorRT-LLM

…