Research Engineer Pre-Training

You develop large-scale foundation models for financial market prediction. You own the training stack by building fault-tolerant infrastructure across GPUs and TPUs, engineering high-throughput data pipelines, designing custom kernels, co-designing model architectures, and advancing mixed-precision training and model parallelism.

Responsibilities

• Build fault-tolerant training infrastructure across GPUs and TPUs
• Engineer high-throughput data pipelines
• Design custom kernels
• Co-design model architectures with researchers
• Develop mixed-precision training approaches
• Develop model-parallel training approaches
• Improve pre-training efficiency and performance

Requirements

• Significant measurable performance improvements in large-scale distributed training
• Published research in efficient training methods, scaling laws, architectures, or ML systems
• Background in numerical computing, HPC, or distributed systems
• Familiarity with GPUs, TPUs, NVLink, InfiniBand, Kubernetes, Slurm, and operating system internals
• Python
• Deep learning frameworks such as PyTorch or JAX
• Advanced degree in computer science, machine learning, physics, mathematics, or related field, or equivalent frontier-lab experience
• CUDA kernel development preferred
• Triton, Pallas, CuTe, PyTorch internals, JAX internals, XLA optimization, or hardware acceleration preferred
• Reinforcement learning, post-training, or fine-tuning knowledge preferred
• Financial markets or trading knowledge preferred

Benefits

• Discretionary bonus eligibility
• Medical, dental, and vision insurance
• HSA, FSA, and Dependent Care options
• Employer-paid group term life and AD&D insurance
• Voluntary life and AD&D insurance
• Paid vacation and paid holidays
• Retirement plan with employer match
• Paid parental leave
• Wellness programs