Machine Learning Researcher - Research & Development
Engineering · London · Full-time · On-site
Annual base salary: £150,000 – £180,000 GBP
Design and train machine learning models that extract signal from market data.
Role description
Westren Capital is a London-based proprietary trading firm specialising in digital asset markets. We are committed to rigorous, first-principles research at the intersection of quantitative finance and machine learning, bringing together talent from Mathematics, Physics, and Computer Science to translate cutting-edge research into actionable signals across global markets.
Our culture is built around intellectual honesty, independence of thought, and a deep respect for evidence over narrative. Collaboration is not ornamental here, it is structural - the best ideas tend to emerge where disciplines overlap and perspectives collide.
This role sits within our AI team, a focused R&D group of quantitative researchers, engineers, and ML practitioners working on frontier problems in representation learning and large-scale modelling applied to markets.
Responsibilities
Design, train, and evaluate machine learning models aimed at extracting signal from unstructured and structured market data
Work closely with traders and quant researchers to understand real constraints, not idealised ones, before building anything
Take models from notebook to production, which means caring about latency, robustness, and failure modes as much as accuracy
Run rigorous ablations and out-of-sample validation; a model that only works in-sample is a warning sign, not a result
Stay current with the ML research landscape and bring back what's actually useful, filtering out what's only impressive
Contribute to shared infrastructure and research workflows that make the next project faster than this one
Requirements
Requirements:
Strong grounding in ML fundamentals, with practical experience in modern architectures such as transformers or related approaches
Comfortable writing production-quality code in Python and/or C++, familiar with a major framework such as PyTorch, TensorFlow, or JAX
Evidence of taking a model from idea to something that worked in a real system, not just a benchmark leaderboard
Curiosity and range matter more than narrow specialism; a working sense of markets, or a willingness to build one fast, is expected
Experience with distributed training, HPC environments, or GPU-level optimisation (CUDA/ROCm) is a strong plus
Comfortable operating with judgement in ambiguous territory, and willing to let that judgement bend when the evidence says otherwise