AI Inference Engineer

Our client, an early-stage AI research lab developing brain-inspired approaches to more efficient AI, is hiring a Research Scientist / Research Engineer focused on inference, making models faster, smaller and more energy efficient, with direct influence on the company’s core technology.
Focus areas (one or more):
Inference profiling and cost modelling (compute, memory capacity and bandwidth, cache growth, latency, throughput, energy)
Efficient inference methods (quantisation, sparsity, pruning, distillation, compression, low-precision computation)
Inference optimisation (KV cache management, speculative decoding, batching, efficient attention, vLLM/TensorRT-LLM/SGLang/llama.cpp)
GPU programming and performance engineering (CUDA, Triton, custom kernels, profiling)
Non-Transformer and brain-inspired architectures (state-space, recurrent, associative-memory, long-context, spiking and neuromorphic models)
Hardware-aware ML (roofline analysis, on-device and edge inference, energy modelling)
Requirements:
Bachelor’s, Master’s or PhD in Maths, CS, Physics, Engineering, Neuroscience or a closely related quantitative field
Strong understanding of modern deep learning and LLM architectures, and of what makes them expensive to run (attention, KV caching, autoregressive decoding)
Strong Python and PyTorch/JAX experience, with hands-on experience profiling and optimising model performance on GPUs, accelerators or CPUs
Strong foundations in several of: linear algebra, probability, optimisation, numerical methods, information theory, dynamical systems or statistical learning
Rigorous experimental mindset, with the ability to move between mathematical reasoning, experimentation and implementation
Research publications, open-source contributions or substantial independent projects valued (industry experience not essential)
Location: London. Full-time, permanent.
If this role sounds interesting to you, please reach out to Charles Duran.