Ai Software Engineer

In this role you will scale novel AI ideas into real systems that run on custom silicon, enabling ultra-efficient inference. You’ll work across model scaling, efficient inference, and non-Transformer architectures to bridge research and production. The position sits in an early-stage AI lab in the UK, collaborating closely with researchers and engineers to translate concepts into systems-level impact. You will tackle hard problems at the intersection of maths, architecture, data, and hardware, shaping how future AI models execute on specialized hardware.

• Scale models from proofs of concept to broader model zoo across sizes, tasks and datasets
• Advance efficient inference through techniques like quantisation, low-bit pipelines, KV-cache compression, and speculative decoding
• Work with non-Transformer model families (state-space, diffusion, sparse/stateful systems)
• Perform kernel- and memory-level optimisations with hardware-aware ML considerations
• Develop and improve evaluation infrastructure (benchmarks, ablation tooling, reference implementations)
• Conduct systems integration work across SDK, compiler, runtime, and hardware abstraction layers
• Translate research ideas into robust systems and software that operate close to hardware boundaries

• 3 to 5 years of hands-on AI/ML engineering or applied AI/ML research experience
• Direct experience with model scaling and scaling-related diagnostics across maths, architecture, data, and systems layers
• Knowledge of several efficient-inference techniques and familiarity with at least one non-Transformer model family
• Strong Python fundamentals; C++ is a plus
• Comfortable in an early-stage, research-led environment without a fixed spec

• problem-solving orientation
• collaboration across research and engineering
• adaptability in ambiguous, early-stage settings
• model scaling
• efficient-inference techniques
• non-Transformer model families