AI Research Engineer, Pre-Training

Trading Interview; New York (NY); On-site; USD 250,000 - 300,000; Full time

Job summary

Hudson River Trading is seeking an AI Research Engineer (Pre-training) to join the HAIL team. You will work on large-scale model training, kernel development, data loading, and parallelism to push the frontier of foundation models for markets.

You will collaborate with researchers to co-design and improve models and research agendas, with access to multiple GPU clusters and a high GPU-to-researcher ratio, making a direct business impact.

Qualifications

• Two+ years of experience building deep learning systems in any domain.
• Experience translating methods across domains is valued.
• LLM experience is valuable but not required.

Responsibilities

• Improve all aspects of large-scale model training, including kernel development, data loading, and fault-tolerance.
• Collaborate with researchers to co-design and improve models and research agenda.

Skills

CUDA / Triton kernel development

Deep learning systems (2+ years)

PyTorch / JAX / XLA

LLM experience (nice to have)

Job description

London, United Kingdom; New York, NY, United States

Job Type Full-time

Posted 2 months ago

The role

Job description

Hudson River Trading (HRT) is seeking an AI Research Engineer (Pre-training) to join the HAIL team. HAIL (HRT AI Labs) is the team at HRT responsible for developing and maintaining our most powerful models, which are used by our trading teams to drive a significant fraction of our trading. We are building and deploying "foundation models for markets", that ingest and train on vast amounts of market and “alternative” data (such as language) to make predictions about future market state.

As a pre-training research engineer on HAIL, you will have a general mandate to improve all aspects of large-scale model training, including but not limited to kernel development, training data loading, parallelism, networking, and fault-tolerance. You will work closely with our researchers to co-design and improve our models, and shape the research agenda.

We have multiple large, modern, and rapidly growing GPU clusters, and we maintain a very high GPU-to-researcher ratio. We are simultaneously pursuing multiple strategies and developing many model types with different purposes, and we are strongly incentivized to squeeze as much as we can out of our systems. Your work will be directly, clearly, and highly impactful on the business, and it will be challenging: this is a field with no easy or obvious solutions.

Qualifications

• Strong engineering skills, especially any of: CUDA/Triton/Pallas/CuTe DSL kernel development, lower-level PyTorch/JAX/XLA development, CUDA Graphs, FPGA/ASIC experience
• Must have two or more years work experience building deep learning systems, for any domain: robotics, biology, chemistry, physics, audio, video, recommendations, etc.
• Experience translating methods between areas of application is highly valued
• LLM experience is valuable, but not necessary
• Finance experience is not required

The estimated base salary range for this position is 250,000 to 300,000 USD per year (or local equivalent). The base pay offered may vary depending on multiple individualized factors, including location, job-related knowledge, skills, and experience. This role will also be eligible for discretionary performance-based bonuses and a competitive benefits package.

Culture

Hudson River Trading (HRT) brings a scientific approach to trading financial products. We have built one of the world's most sophisticated computing environments for research and development. Our researchers are at the forefront of innovation in the world of algorithmic trading.
At HRT we welcome a variety of expertise: mathematics and computer science, physics and engineering, media and tech. We’re a community of self-starters who are motivated by the excitement of being at the cutting edge of automation in every part of our organization- from trading, to business operations, to recruiting and beyond. We value openness and transparency, and celebrate great ideas from HRT veterans and new hires alike. At HRT we’re friends and colleagues - whether we are sharing a meal, playing the latest board game, or writing elegant code. We embrace a culture of togetherness that extends far beyond the walls of our office.
Feel like you belong at HRT? Our goal is to find the best people and bring them together to do great work in a place where everyone is valued. HRT is proud of our diverse staff; we have offices all over the globe and benefit from our varied and unique perspectives. HRT is an equal opportunity employer; so whoever you are we’d love to get to know you.

HRT’s founding partners graduated from Harvard and MIT with degrees in computer science and mathematics, and founded HRT in 2002. Armed with programming skills and sharp quantitative minds, the founders…