Machine Learning Performance Engineer (CUDA)

Fintal Partners; New York (NY); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

401(k) matched up to 50%

Health, dental, vision, disability & &

Life cover

25 days' paid leave

Office breakfast, lunch & snacks

Social events & clubs

Job summary

Fintal Partners is seeking a Machine Learning Performance Engineer to join a flagship AI program in New York. You will shape the ML training and inference stack, optimize pipelines, and collaborate with researchers and engineers to push the performance boundaries within our trading infrastructure.

Based in New York, you’ll work with a diverse team across 40+ nationalities, enjoying generous compensation, comprehensive health coverage, and ample paid leave and office perks to support deep focus

Qualifications

• Strong low-level CUDA programming skills.
• Deep familiarity with frameworks such as PyTorch, JAX or TensorFlow.
• Solid grounding in computer architecture.
• Proficiency in C++ and Python.
• Experience with the JAX ecosystem (XLA, Flax and similar).
• Experience with large-scale distributed training and open-source ML projects.

Responsibilities

• Build robust, scalable training and inference pipelines for deep learning.
• Work inside the internals of open-source deep learning frameworks and extend what they can do.
• Hunt down and remove performance bottlenecks in ML systems.
• Partner closely with researchers and fellow engineers to advance performance.
• Build deep knowledge of how the trading systems work.

Skills

CUDA GPU programming

PyTorch

JAX

TensorFlow

C++

Python

Distributed training

Open-source contributions

Computer architecture

Tools

XLA

Flax

Git

Job description

A leading global trading firm is hiring a Machine Learning Performance Engineer to work on a flagship AI program. You'll have real influence across ML infrastructure, training and inference, work that feeds directly into the firm's futures trading strategies.

The role

• Build robust, scalable training and inference pipelines for deep learning
• Work inside the internals of open-source deep learning frameworks and extend what they can do
• Hunt down and remove performance bottlenecks
• Partner closely with researchers and fellow engineers
• Build deep knowledge of how the trading systems work

What's on offer

• You'll be part of a collaborative, high-performing culture built around curious thinkers and inventive problem-solvers, tackling some of the hardest challenges in financial markets alongside genuinely strong colleagues.
• Colleagues drawn from 40+ nationalities
• Highly competitive compensation
• 401(k) matched up to 50%
• Full health, mental health, dental, vision, disability and life cover
• 25 days' paid leave plus market holidays
• Generous office perks — breakfast, lunch and snacks, social events, clubs and sports leagues

What you'll bring

• Strong low-level GPU programming in CUDA — Tensor Cores, cooperative groups, graphs, warp-level intrinsics
• Deep familiarity with the internals of frameworks such as PyTorch, JAX or TensorFlow
• Solid grounding in computer architecture
• C++ and Python
• The JAX ecosystem (XLA, Flax and similar)
• Large-scale distributed training
• Open-source contributions in ML or data science

About the firm

A major market maker active on 100+ exchanges worldwide, focused on improving markets by adding liquidity, pricing accurately and increasing transparency in all conditions — with the health and efficiency of those markets a priority for every participant.