Quantitative AI Research Scientist — Market Foundations

JPMorganChase; Hong Kong; On-site; HKD 900,000 - 1,500,000; Full time

Job summary

JPMorgan Chase seeks an AI/ML quantitative researcher to lead pre-training of Transformer-based and time-series foundation models on vast market datasets. You will develop representations, tokenization, and distributed training recipes for financial time series, preparing models for alpha generation, pricing, market making, and risk tasks.

You will explore scaling, cross-asset transfer, and robustness, and collaborate with ML infra engineers to build reusable training and deployment components

Qualifications

• Advanced degree (Master's, PhD, or equivalent) in ML, CS, statistics, mathematics, operations research, engineering, or a related quantitative field.
• Demonstrated experience pre-training a large model from scratch (Transformer/LLM/multimodal/time-series). API usage or prompt engineering is not sufficient.
• Experience building large-scale data pipelines and distributed training systems using PyTorch, JAX, or equivalent frameworks.
• Deep knowledge of large-model training and evaluation: optimization, parallelism, mixed precision, checkpointing, experiment design, ablations, and benchmarking.
• Evidence of research/technical quality through successful large-model training, high-impact research, open-source systems, or production deployment.

Responsibilities

• Pre-train Transformer-based and time-series foundation models from scratch using large-scale market, order-book, transaction, and cross-asset datasets.
• Develop data representations, tokenization schemes, self-supervised objectives, model architectures, and distributed training recipes for financial time series.
• Fine-tune and post-train foundation models for alpha generation, pricing, market making, execution, and risk-management tasks.
• Study scaling laws, transfer across instruments and asset classes, regime robustness, data efficiency, and the trade-offs among model quality, inference cost, and latency.
• Design evaluation protocols that connect pre-training metrics to economically meaningful outcomes, including out-of-sample prediction, simulated trading, transaction costs, capacity, and live markouts.
• Build reusable training, checkpointing, evaluation, and model-serving components with ML infrastructure engineers.

Skills

Transformer training

Distributed training

Experiment design

Education

Advanced degree in quantitative field

Tools

PyTorch

JAX

Job description

JPMorgan Chase seeks an AI/ML quantitative researcher to lead pre-training of Transformer-based and time-series foundation models on vast market datasets. You will develop representations, tokenization, and distributed training recipes for financial time series, preparing models for alpha generation, pricing, market making, and risk tasks.

You will explore scaling, cross-asset transfer, and robustness, and collaborate with ML infra engineers to build reusable training and deployment components