AI Research Engineer — Production-Scale ML & HPC

Jump Trading; New York (NY); On-site; USD 270,000 - 360,000; Full time

Job summary

Jump Trading Group seeks exceptional ML researchers to advance financial market models and scalable AI systems. You will join a collaborative team of researchers, quants, and engineers who work together to push scientific boundaries and deploy impactful technology.

Expect to publish and prototype novel ML methods, optimize training pipelines for HPC, and integrate latency-sensitive models into production using Python, C++, CUDA, and other low-level GPU languages.

Qualifications

• Strong publication record at ICML, ICLR, NeurIPS, or equivalent.
• Strong ML background with exposure to modern deep learning techniques and/or language modeling architectures (e.g. transformers, SSMs)
• Proficiency in Python and/or C++.
• Experience with PyTorch, JAX, or TensorFlow.
• Collaborative mindset and ability to work in a research/trading environment.

Responsibilities

• Apply state-of-the-art ML techniques to complex and challenging domains.
• Work closely with researchers and quants to build flexible and reusable frameworks for financial ML.
• Optimize training pipelines to make efficient use of HPC resources.
• Integrate ML models into production systems where latency matters.
• Work across languages: C/C++, Python, CUDA and other low-level GPU languages.
• Build large-scale ML systems that are observable, performant, and flexible; reduce iteration cycle time.
• Other duties as assigned or needed.

Skills

Python

C++

CUDA

Machine learning

Deep learning

PyTorch

JAX

TensorFlow

Communication

Tools

CUDA toolkit

Job description

Jump Trading Group seeks exceptional ML researchers to advance financial market models and scalable AI systems. You will join a collaborative team of researchers, quants, and engineers who work together to push scientific boundaries and deploy impactful technology.

Expect to publish and prototype novel ML methods, optimize training pipelines for HPC, and integrate latency-sensitive models into production using Python, C++, CUDA, and other low-level GPU languages.