Senior Data Scientist

The Role

Our current work includes:

• Agentic pipelines — multi-step LLM systems with tool use, planning, and self-evaluation that automate complex marketing workflows end to end.
• Domain-adapted foundation models — fine-tuning open-weight LLMs (LoRA, RLHF, distillation) on proprietary WPP data for tasks like audience segmentation, creative scoring, and brand-safety classification.
• Retrieval-augmented generation — production RAG systems over large proprietary corpora (embedding models, vector indices, re-ranking) that serve real-time answers to client queries.
• Classical ML at scale — gradient-boosted models, causal inference pipelines, and recommendation engines that run alongside LLM components in hybrid architectures.

You will be the technical lead for projects across these workstreams: scoping the problem, choosing the modelling approach, building the training and evaluation infrastructure, shipping to production, and iterating based on live metrics. You are not handing off a notebook to an engineering team — you ship what you build.

What You'll Do

• Design and run training pipelines — data curation, model selection, hyperparameter search, ablation studies — and be accountable for model quality on live traffic.
• Build and maintain production inference services (latency budgets, batching strategies, quantisation, monitoring) that serve WPP's global client base.
• Architect agentic AI systems: define tool schemas, orchestration logic, evaluation criteria, and failure modes for multi-step LLM workflows.
• Work across the stack when needed — write the data pipeline, train the model, build the evaluation harness, deploy the service, and debug it when metrics drift.
• Set technical direction for your workstream: write design docs, make build-vs-buy decisions, and defend your approach with evidence.
• Mentor and set the quality standards for junior scientists.

What We're Looking For

• 5+ years shipping ML models to production — you've dealt with data drift, silent failures, retraining cadences, and the gap between offline metrics and business outcomes.
• Deep, demonstrable expertise in at least one of: NLP/LLMs, computer vision, recommender systems, or causal inference.
• Hands-on experience with LLM fine-tuning (LoRA, RLHF, DPO) or building LLM-powered systems (agents, RAG, structured generation).
• Strong software engineering habits — version control, testing, code review, CI/CD.
• Comfort with ambiguity. Many of our problems don't have a known-good solution.
• Clear communication — you can write a one-page design doc that is useful for both product managers and staff engineers.

Nice to Have

• Experience with multimodal models.
• Background in marketing technology, ad tech, audience modelling, or media mix modelling.
• Publications at top venues (NeurIPS, ICML, ACL, CVPR) or meaningful open-source contributions.

What We Offer

• Benefits - enhanced pension, life assurance, income protection, private healthcare;
• Remote working - café, bedroom, beach - wherever works;
• Truly flexible working hours - school pick up, volunteering, gym;
• Generous Leave - 27 days holiday plus bank holidays and enhanced family leave;
• Annual bonus - when [Employer hidden — sign up to reveal] does well, we all do well;
• Impactful projects - focus on bringing meaningful social and environmental change;
• People oriented culture - wellbeing is a priority, as is being a nice person;
• Transparent and open culture - you will be heard;
• Development - focus on bringing the best out of each other;

[Employer hidden — sign up to reveal] is home to some of the brightest minds in AI and if you’re looking to join a company who not only values autonomy and freedom, but embraces a culture of inclusion and warmth, we’d love to hear from you.