Gensyn · 2 days ago

GPU Software Engineer

United States

Full-time

Remote

Mid Level

Maximize your interview chances

BlockchainCryptocurrency

Insider Connection @Gensyn

Discover valuable connections within the company who might provide insights and potential referrals.
Get 3x more responses when you reach out via email instead of LinkedIn.

Responsibilities

Implement core GPU components for distributed ML training wrapped in a decentralised protocol

Develop performant GPU kernels and compute infrastructure - from the framework level (e.g. PyTorch) down to IR for training, with a strong emphasis on reproducibility in multi-GPU distributed training environments

Design novel algorithms - with a focus on numerical properties and stable compute flows, optimised for modern cryptographic systems

Qualification

Find out how your skills align with this job's requirements. If anything seems off, you can easily click on the tags to select or unselect skills to reflect your actual expertise.

GPU Software EngineeringCUDAPyTorchDeep LearningDistributed GPU ComputeNumerical StabilityMachine Learning FrameworksHeterogeneous System ArchitectureComputer ArchitectureCompiler TechnologyDeep Learning Compiler FrameworksDistributed Training InfrastructureOpen-source Contributions

Required

Strong software engineering skills - with substantial experience as a practising software engineer and significant contributions to shipping production-level code

Hands on experience in distributed GPU compute environments: Writing GPU Kernels (e.g. CUDA, PTX, MPX/MLX, IR); and/or Implementing low-level GPU-specific optimizations for performance, numerical stability and determinism

In-depth understanding of deep learning - including recent architectural trends, training fundamentals, and practical experience with machine learning frameworks and their internal mechanics (e.g., PyTorch, TensorFlow, JAX)

Preferred

Deep understanding of heterogenous system architecture

Experience in a venture backed start-up environment

Open-source contributions to high-performance GPU codebases

Strong understanding of computer architecture - with expertise in specialised architectures for training neural networks, including Intel Xeon CPUs, GPUs, TPUs, and custom accelerators, as well as heterogeneous systems combining these components

Solid foundation in compiler technology - with a working knowledge of traditional compilers (e.g., LLVM, GCC) and graph traversal algorithms

Experience with deep learning compiler frameworks - such as TVM, MLIR, TensorComprehensions, Triton, and JAX

Experience working with distributed training infrastructure and software development