AI Research Scientist, Quantitative Trading & Time-Series

JPMorganChase; Singapore; On-site; SGD 180,000 - 240,000; Full time

Job summary

J.P. Morgan's AI Market Lab in Singapore seeks an AI/ML quantitative researcher to build Transformer-based and time-series foundation models for trading across FX, rates, commodities, and equities.

You will work on pre-training from scratch, rigorous evaluation, and deployment within a fast-paced research-to-production pipeline. You should have a strong research background, hands-on model training experience, and the ability to operate under real-market constraints, including latency and data

Qualifications

• Advanced degree in ML, CS, or quantitative field.
• At least 2 years of hands-on relevant experience.
• Experience pre-training a large model from scratch.
• Experience with PyTorch or JAX and distributed training.
• Proven research quality through publications or deployments.

Responsibilities

• Pre-train Transformer- and time-series foundation models from scratch on large market datasets.
• Develop data representations, tokenization, self-supervised objectives, and distributed training.
• Fine-tune models for alpha, pricing, market making, execution, and risk management.
• Study scaling laws, cross-instrument transfer, regime robustness, data efficiency, latency trade-offs.
• Design evaluation protocols linking pre-training metrics to live trading outcomes.
• Collaborate with ML infra engineers to build training, checkpointing, and model serving tools.

Skills

Advanced degree

2+ years experience

Pre-training large models

PyTorch

JAX

Distributed training

Education

Master's or PhD

Tools

PyTorch

JAX

Job description

J.P. Morgan's AI Market Lab in Singapore seeks an AI/ML quantitative researcher to build Transformer-based and time-series foundation models for trading across FX, rates, commodities, and equities.

You will work on pre-training from scratch, rigorous evaluation, and deployment within a fast-paced research-to-production pipeline. You should have a strong research background, hands-on model training experience, and the ability to operate under real-market constraints, including latency and data