Train, evaluate, and iterate on ML models and agentic systems for customer feedback, including owning our custom fine-tuning pipelines. Run experiments end-to-end, track results rigorously, and make clear recommendations on what to ship, iterate, or retire.
Build and maintain LLM-powered features: retrieval pipelines, reranking systems, insight agents, data mining agents, and automated taxonomy generation.
Design and run robust evaluation frameworks: build test sets, define metrics, evaluate non-deterministic systems, handle class imbalance, and automate checkpoint comparisons.
Improve and extend semantic search and retrieval, evolving from embedding-based approaches toward more advanced methods.
Write production-quality code and collaborate closely with Engineering on productionisation, model serving, data pipelines, and monitoring.
Work with Product and Commercial teams to translate business needs into practical ML solutions, and support client evaluations and accuracy benchmarking.
Mentor team members, review code and research, and bring relevant advances from the literature into the product.
Requirements
A deep working knowledge of transformer architectures.
Strong PyTorch skills, with the ability to write custom training loops, modify model architectures, and debug issues at the tensor level. Ideally, experience with parameter-efficient fine-tuning techniques such as LoRA
Extensive experience working with large-scale, messy real-world text data, including classification, extraction, embeddings, re-rankers, clustering, and search.
Experience in instruction fine-tuning and serving language models, familiarity with frameworks such as vLLM, DeepSpeed, or similar tools
A solid grounding in classical ML and statistics, and the judgement to choose simpler methods when they’re the right solution.
Practical experience building with GenAI and agentic patterns.
Excellent communication skills and confidence translating complex technical concepts for non-technical audiences (and vice versa!).
Technical curiosity and a keen interest in AI – a love of experimenting to make the most of available technology.
High ownership and initiative, with the ability to identify problems, prioritise effectively, and drive solutions forward.
Tech Stack
PyTorch
Benefits
Monthly Health & Wellness budget, increasing with length of service
Annual Learning and Development budget, increasing with length of service
Flexible working in a choice-first environment
we trust the way you want to work!
Work From Home Allowance
25 Holiday Days + your local bank holidays, plus an extra day for every year of service
Your birthday off
Enhanced Family Leave (UK Only), Fertility Leave, and Neonatal Leave
Optional Healthcare Plan
Life & income protection (Location dependent)
Employee Assistance Programme (UK Only)
The opportunity to share in the company’s success through options
If you’re in London, a dog-friendly office with great classes, events, and a rooftop terrace