The RoleWe are looking for an Open Source Software Engineer to build and optimize machine learning systems that integrate with AMD hardware. You will work at the intersection of open source AI software, GPU systems, and high-performance computing, helping bring cutting-edge AI workloads to AMD platforms.
You will design, implement, test, and ship open source integrations with widely used AI systems and frameworks. The role requires a strong understanding of modern LLM inference and compute, GPU kernel optimization, and the realities of developing software in collaborative open source communities.
This is an engineering role for someone who enjoys working close to the hardware while also contributing upstream to the software ecosystem.
The PersonYou are a systems-oriented software engineer who understands how modern ML workloads map onto GPUs. You are comfortable moving between high-level ML frameworks and low-level kernels, and you enjoy investigating performance bottlenecks and turning them into measurable improvements.
You have:
- Strong familiarity with LLM systems, inference workloads, compute, and optimization.
- Experience optimizing GPU kernels, particularly for AMD architectures.
- The ability to learn and adapt across different kernel programming languages and programming models.
- A strong understanding of kernel fusion, GPU execution, and pipelined execution.
- An understanding of performance tradeoffs across compute, memory bandwidth, synchronization, and data movement.
- Familiarity with open source development practices, including code review, automated testing, CI, and upstream contribution.
- A willingness to work across layers of the stack, from ML models and frameworks down to GPU kernels and hardware-specific optimizations.
- Knowledge of graphics processor pipelines and experience with quantized data layouts or model quantization techniques are valuable additions.
Key Responsibilities- Develop and maintain open source integrations between ML/AI software and AMD hardware.
- Optimize kernels and critical execution paths for AMD GPUs.
- Profile workloads, identify performance bottlenecks, and implement measurable optimizations.
- Develop and optimize kernel fusions and GPU execution pipelines.
- Adapt implementations across different kernel programming languages and GPU programming environments.
- Work with LLM inference systems and understand how model architecture, data layout, quantization, and execution strategy affect performance.
- Contribute high-quality code upstream to relevant open source projects.
- Write comprehensive unit, integration, and performance tests.
- Build and maintain automated testing and validation infrastructure.
- Participate in code reviews and collaborate with maintainers and contributors across the open source community.
- Investigate and resolve correctness and performance issues across the software and hardware stack.
- Document implementations, performance characteristics, and technical decisions for other developers.
What Success looks like
In this role, you will help make advanced AI workloads
fast, reliable, and broadly accessible on AMD hardware. Success means not only producing high-performance implementations, but contributing them in a form that can be reviewed, tested, maintained, and adopted by the broader open source community.
Preferred ExperienceExperience with one or more of the following technologies is a strong plus:
- vLLM
- SGLang
- llama.cpp
- ROCm
- CUDA
- Vulkan / SPIR-V
- Additional valuable experience includes:
- GPU kernel development and optimization.
- AMD GPU architecture and performance analysis.
- LLM inference and serving systems.
- Kernel fusion and pipelined GPU execution.
- GPU and graphics hardware pipelines.
- Quantized models, quantization algorithms, and quantized data layouts.
- Multiple GPU programming languages or kernel DSLs.
- Performance profiling and benchmarking.
- Contributing to large, active open source projects.
- Open source CI, automated testing, code review, and release processes.
Experience with CUDA is useful even when the primary target is AMD hardware; familiarity with one GPU programming environment should translate into an ability to learn and work effectively in others.
Preferred Academic CredentialA Bachelor's or Master's degree in Computer Science, Computer Engineering, Electrical Engineering, or a related technical field is preferred.
Equivalent professional experience demonstrating strong systems programming, GPU computing, machine learning infrastructure, and open source engineering expertise will also be considered.
#LI-G11
#LI-HYBRID
Benefits offered are described: AMD benefits at a glance.
