ML Systems Engineer

Final is a world leader in trading algorithms and trade execution technologies development. Our multi-disciplinary teams have built a unique and highly successful machine-learning–based HFT platform that consistently delivers exceptional results. In a world increasingly dominated by learning machines and artificial intelligence, we at Final are especially proud of our humans. Our elite team of exceptional individuals is the soul of our company, and we prioritize providing them with a professionally fulfilling environment that supports a healthy work–life balance. We encourage our employees to pursue their passions outside of work and are proud to offer a variety of opportunities, resources, and an agile environment that promotes their well-being. We are looking for a Software Engineer specializing in ML Systems Engineering to join Final’s Development department. While vision and language models have become increasingly commoditized, Final’s proprietary deep learning models are unique, fast-evolving, and deployed in live trading across the world’s most efficient and sophisticated financial markets. Operating in this environment presents distinct scaling challenges and continuous opportunities for optimization. Success in this role requires first-principles thinking and a deep understanding of the engineering trade-offs behind high-performance DL systems. This is a pivotal role within Final’s engineering organization. You will work closely with researchers and engineers across the company, running deep learning models on massive compute clusters and adapting them for production serving under strict and non-trivial constraints. Requirements B.Sc. with honors in CS/EE/Math/Physics, or a related field from a top-tier university 5+ years of hands-on experience building and deploying large-scale deep learning systems in production Advanced proficiency in PyTorch/TensorFlow Preferred Qualifications : Proficiency in Python/C/C++ Deep, working knowledge of PyTorch internals Performance profiling and optimization of deep learning workloads Orchestrating and optimizing large-scale distributed training (hundreds to thousands of GPUs) Optimizing model serving and inference pipelines (quantization, distillation, compilation, memory optimization, etc.) Training and scaling state-of-the-art vision, language, or diffusion models Implementing custom CUDA/Triton kernels