About this role
Job title: Research Engineer, Pre-Training
About the Role Jump Trading Group is building the future of ML-powered trading and foundation model research. The Pre-Training Engineer will lead development of massive-scale foundation models, owning the training stack from infrastructure to data pipelines, and collaborating with researchers to push the boundaries of pre-training at scale.
What You'll Do
- Own and drive the entire training stack: building fault-tolerant infrastructure that scales across thousands of GPUs and TPUs with near-linear performance.
- Engineer data pipelines that stream terabytes per second as models train on petabytes of data from global markets.
- Design custom kernels to unlock 10x efficiency gains and co-design novel architectures with researchers.
- Pioneering cutting-edge approaches to mixed-precision training and model parallelism with the latest hardware.
- Work closely with researchers to push the boundaries of pre-training at scale and translate research gains into live trading improvements.
- Other duties as assigned or needed.
What We're Looking For
- Expertise and track record of significant, measurable performance improvements in large-scale distributed training (MFU, throughput, convergence, cost-per-token).
- Published research in efficient training methods, scaling laws, architectures, or systems for ML.
- Background in numerical computing, HPC, or distributed systems, including familiarity with GPUs/TPUs, high-performance networking (NVLink/InfiniBand), Kubernetes/Slurm, and OS internals.
- Expertise in Python and deep experience with modern deep learning frameworks (PyTorch and/or JAX).
- Advanced degree (MS or PhD) in Computer Science, Machine Learning, Physics, Mathematics, or a related quantitative field, or equivalent industry experience at a frontier lab.
- Ability to balance ambitious research goals with practical engineering constraints.
- Strong problem-solving skills, results orientation, and excellent collaborative communication.
- Reliable and predictable availability.
Nice to Have
- Expertise in CUDA kernel development, Triton/Pallas/CuTe DSLs, PyTorch/JAX internals, XLA optimization, or hardware acceleration (FPGA/ASIC).
- Knowledge of reinforcement learning, post-training, or fine-tuning techniques.
- Knowledge of financial markets or trading.
Compensation & Benefits
- Annual Base Salary Range: $300,000–$350,000 USD
- 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 & AD&D insurance
- Paid vacation plus paid holidays
- Retirement plan with employer match
- Paid parental leave
- Wellness Programs