Machine Learning Engineer - Hybrid

Senior LLM Engineer London | Hybrid (2 days per week on-site, Bank area)
My client is a long-established consultancy that originally built its reputation in data engineering and wider data transformation, and over the past few years has grown a strong applied AI practice. They tend to work with UK organisations where data security, reliability and explainability actually matter, so think regulated, risk-aware, or operationally critical environments rather than hype-driven MVPs.

The current AI/ML group is around eight senior engineers, with backing to grow sensibly (not a hiring spree). This role sits within that team and is focused on designing and deploying LLM systems that make it into production, are cost-efficient, and work reliably in the real world.

Owning the design and build of LLM-based systems end-to-end
Fine-tuning and adapting models (LoRA, instruction tuning, PEFT etc) rather than just prompt-engineering
Building RAG workflows, embedding strategies, memory layers and domain grounding
Working out how to measure output quality, reduce hallucination risk and improve robustness
Optimising inference performance (quantisation, distillation, pruning, batching, caching)
Deploying models into production environments where latency, cost and privacy are real constraints
Strong Python and experience with PyTorch / Hugging Face / similar tooling
Experience deploying models into production (not just training them)
Comfortable operating in a hybrid environment and speaking with stakeholders where required
Someone who enjoys solving real problems end-to-end: from understanding the domain designing the approach shipping the implementation
No relocation support youll need to already be UK-based and able to be in London (Bank area) twice a week
This is a senior role typically someone with 3+ years experience in ML/AI roles, ideally from a data science or applied ML background

27 days annual leave + bank holidays
4-week paid sabbatical after 4 years tenure
Regular team meet-ups, workshops, and knowledge-sharing sessions
Clear progression as the AI group continues to scale