Senior Research Engineer (AI Inference)

Role: Staff Software Engineer (Rust & AI Inference)

Location: London (2-3 days a week onsite)

Compensation: Up to £160,000 + equity

We're supporting an emerging deep tech company building high-performance infrastructure to support AI model deployment at scale. They're focused on enabling enterprises to run advanced AI systems in production, with a strong emphasis on privacy, performance, and full-stack control.

This is a hands-on senior IC role for an engineer who thrives in low-latency, compute-intensive environments, and enjoys mentoring others. You'll play a key role in shaping the core inference layer powering complex real-time AI workflows.

If you're excited by the challenge of optimising distributed systems, designing for reliability, and scaling cutting-edge AI applications, this could be the role for you.

You'll be:

• Building scalable, low-latency LLM inference infrastructure
• Optimising performance with caching, quantisation, and speculative decoding
• Contributing to core systems in Rust (we are happy for individuals to lean this on the job)
• Designing distributed GPU orchestration and inference servers
• Mentoring engineers and influencing technical direction across the team

You should bring:

• 5+ years software engineering experience, with deep backend/system-level experience
• Strong coding skills in a typed systems languages like C++ / Go / Rust (would also consider Python)
• Familiarity with Kubernetes, and cloud infra
• A strong engineering mindset with a bias for clean abstractions, reliability, and performance

Bonus points for:

• Experience deploying open-source LLMs or VLMs in production
• Experience with ML inference systems, PyTorch, Triton, or CUDA kernels
• Background in document intelligence, enterprise search, or NLP pipelines
• Prior exposure to multi-agent systems or complex orchestration workflows

This is a high-impact, high-autonomy role in a technically elite team that's pushing the boundaries of what's possible in AI infrastructure.