AI Research Engineer, Pre-Training

Overview

As an AI Research Engineer (Pre-training) on the HAIL team, you will advance large-scale model training for market-focused foundation models. You’ll collaborate with researchers to co-design, optimize training systems, and scale GPUs across busy clusters. This role blends deep engineering with research to push performance, fault tolerance, data pipelines, and parallelism in a demanding, finance-oriented environment. You’ll impact trading outcomes by improving the core infrastructure that drives state-of-the-art models.

Pay / Benefits

• discretionary performance-based bonuses
• competitive benefits package
• salary range 250000-300000 USD per year
• global offices
• equal opportunity employer

Responsibilities

• Improve large-scale model training end-to-end (kernel development, data loading, parallelism, networking, fault-tolerance)
• Collaborate with researchers to co-design and enhance models and research agenda
• Manage and optimize GPU clusters for high-performance training
• Advance systems capabilities across kernel, framework, and toolchains (CUDA, PyTorch/JAX/XLA, CUDA Graphs)
• Develop and maintain scalable training pipelines and fault-tolerant workflows
• Contribute to a culture of experimentation and rapid iteration to squeeze performance from systems

Key requirements

• Strong engineering skills in CUDA/Triton/Pallas/CuTe DSL kernel development
• Lower-level PyTorch/JAX/XLA development experience
• CUDA Graphs or FPGA/ASIC experience
• Two+ years building deep learning systems across domains
• Experience translating methods between applications; LLM experience valued but not required
• Finance experience not required

• Collaborative mindset
• Strong problem-solving and cross-disciplinary communication
• Adaptability in a fast-growing environment
• CUDA/Triton/Pallas/CuTe DSL kernel development
• Lower-level PyTorch/JAX/XLA development
• CUDA Graphs

…