Quantitative Researcher (Machine Learning)

Thurn Partners; Greater London; On-site; GBP 150,000 - 230,000; Full time

Job summary

Thurn Partners in London is building a specialist ML team focused on large-scale sequential modelling for alpha generation. The role involves transformer-based and other deep learning models applied to market data, from representation learning to live forecasts feeding live strategies.

You will scale training on large GPU clusters, develop robust data pipelines for order book and bar data, and translate model outputs into tradable signals while monitoring PnL in production.

Qualifications

• Expertise in modern deep learning, with hands-on experience building and training transformer-based or other sequential models at scale.
• Demonstrable experience applying ML to financial time series in a systematic trading setting, ideally across equities and/or futures at mid-frequency horizons.
• Proficiency in Python and a major deep learning framework (PyTorch or JAX). C++ and experience with distributed training or high-performance computing environments are a plus.

Responsibilities

• Design, train, and evaluate transformer-based and other sequential deep learning models to forecast returns and related market quantities across equities and futures.
• Build and own the data and feature pipeline for market data tokenisation and cross-asset context.
• Develop robust training and validation frameworks for noisy financial data, including leakage-safe cross-validation and walk-forward testing.
• Scale model training and inference on large GPU clusters, optimizing throughput, latency, and stability with ML infra teams.
• Turn model outputs into tradable signals, collaborating with portfolio construction and execution teams.
• Iterate on live performance: monitor model behaviour and PnL in production, diagnose alpha decay, and retrain over time.

Skills

Transformer models

Deep learning

Python

PyTorch/JAX

Distributed training

Tools

C++

Job description

Company: A leading quantitative proprietary HFT firm expanding into mid-frequency strategies across global equities and futures markets.

Location: London

The role: The firm is building a specialist machine learning team focused on large-scale sequential modelling for alpha generation. The team develops transformer-based and other deep learning models on market data, from representation learning through to forecasts that feed directly into live, risk-bearing equities and futures strategies.

Responsibilities

• Design, train, and evaluate transformer-based and other sequential deep learning models (e.g. attention architectures, state-space models, recurrent networks) to forecast returns and related market quantities across equities and futures.
• Build and own the data and feature pipeline: how order book, trade, and bar-level data are tokenised and represented, how cross-asset and cross-sectional context is brought in, and how the models cope with non-stationarity and regime change.
• Develop rigorous training and validation frameworks for noisy, low signal-to-noise financial data, including leakage-safe cross-validation, walk-forward testing, and robust out-of-sample evaluation.
• Scale model training and inference on large GPU clusters, optimising for throughput, latency, and stability, working closely with ML infrastructure and engineering teams.
• Turn model outputs into tradable signals, working with portfolio construction and execution teams on turnover, transaction costs, and capacity.
• Iterate on live performance: monitor model behaviour and PnL in production, diagnose alpha decay and drift, and retrain and refine models over time.

Requirements

• Expertise in modern deep learning, with hands-on experience building and training transformer-based or other sequential models at scale (e.g. attention mechanisms, self-supervised pre-training, state-space models).
• Demonstrable experience applying ML to financial time series in a systematic trading setting, ideally across equities and/or futures at mid-frequency horizons.
• Proficiency in Python and a major deep learning framework (PyTorch or JAX). C++ and experience with distributed training or high-performance computing environments are a plus.
• A track record of taking research into production and generating live PnL is highly valued.
• Publications at leading ML venues (NeurIPS, ICML, ICLR) or experience at a top AI research lab are a plus. Experience with market microstructure or transaction cost modelling is preferred.