ML/AI Software Engineer - Triton GPU Kernel Optimization

WHAT YOU DO AT AMD CHANGES EVERYTHING We care deeply about transforming lives with AMD technology to enrich our industry, our communities, and the world. Our mission is to build great products that accelerate next-generation computing experiences – the building blocks for the data center, artificial intelligence, PCs, gaming and embedded. Underpinning our mission is the AMD culture. We push the limits of innovation to solve the world’s most important challenges. We strive for execution excellence while being direct, humble, collaborative, and inclusive of diverse perspectives. AMD together we advance_ THE PERSON: You have strong technical and analytical skills and experience in C/C++ AI development and GPU programming in Triton, CUDA or HIP. You are a capable problem solver and also a technical leader to define goals and scope and drive development effort. You possess good communication skills and are enthusiastic about AI technologies and strongly motivated to enable our customers with the best feature-rich efficient solutions. THE ROLE: We are looking for an experienced ML/AI software engineer with deep expertise in GPU kernel development and building high‑performance primitives for training and inference. You’ll design, implement, and optimize custom Triton kernels for core ML workloads integrate them into our frameworks and services, and drive end‑to‑end performance on AMD ROCm platform across AMD Radeon and Ryzen product families. KEY RESPONSIBILITIES: Design, implement, and maintain Triton GPU kernels for state‑of‑the‑art ML workloads, with a focus on fusion, tiling, vectorization, and memory‑hierarchy optimization for AMD RDNA GPU architectures. Analyze performance using profiling tools; identify bottlenecks in memory hierarchy, tensor core/matrix core utilization, warp/wavefront scheduling, and synchronization Optimize kernels across problem shapes, batch sizes, and sequence lengths; implement autotuning strategies and performance heuristics. Integrate custom kernels with PyTorch (torch.compile, torch._inductor), JAX, or other frameworks; develop Python/C++ bindings and align with runtime/graph compilers. Track and contribute to upstream Triton and related compiler ecosystems; propose enhancements aligned with our workloads. PREFERRED EXPERIENCE: Triton GPU kernel development and optimization experience CUDA or HIP kernel development experience (porting, performance tuning, and feature enablement) Experience optimizing kernels on AMD GPUs and familiarity with the ROCm software stack (HIP runtime, rocBLAS, MIOpen, rocWMMA, RCCL, etc.) is a plus Familiarity with compiler internals and IRs (LLVM, MLIR) and codegen for GPUs Strong background in linear algebra, convolution algorithms, attention mechanisms, or other core ML primitives Experience integrating kernels with PyTorch, JAX, or TensorFlow, including custom ops/extensions Knowledge of quantization (INT8/FP8), mixed precision, custom dtypes, and numerics (stability, error analysis). Experience with LLM, diffusion and MoE workloads: FlashAttention‑style kernels, paged attention, grouped‑query attention, rotary embeddings, fused MLPs. Contributions to open-source projects in Triton, ROCm, or related GPU/ML ecosystems are a plus ACADEMIC CREDENTIALS: Bachelor’s, Master, or PhD in Computer Science, Electrical Engineering or relevant fields. LOCATION: Belgrade or Nis (Serbia) #LI-CM1 #LI-HYBRID Benefits offered are described: AMD benefits at a glance. AMD does not accept unsolicited resumes from headhunters, recruitment agencies, or fee-based recruitment services. AMD and its subsidiaries are equal opportunity, inclusive employers and will consider all applicants without regard to age, ancestry, color, marital status, medical condition, mental or physical disability, national origin, race, religion, political and/or third-party affiliation, sex, pregnancy, sexual orientation, gender identity, military or veteran status, or any other characteristic protected by law. We encourage applications from all qualified candidates and will accommodate applicants’ needs under the respective laws throughout all stages of the recruitment and selection process.

Benefits offered are described: AMD benefits at a glance. AMD does not accept unsolicited resumes from headhunters, recruitment agencies, or fee-based recruitment services. AMD and its subsidiaries are equal opportunity, inclusive employers and will consider all applicants without regard to age, ancestry, color, marital status, medical condition, mental or physical disability, national origin, race, religion, political and/or third-party affiliation, sex, pregnancy, sexual orientation, gender identity, military or veteran status, or any other characteristic protected by law. We encourage applications from all qualified candidates and will accommodate applicants’ needs under the respective laws throughout all stages of the recruitment and selection process.

THE PERSON: You have strong technical and analytical skills and experience in C/C++ AI development and GPU programming in Triton, CUDA or HIP. You are a capable problem solver and also a technical leader to define goals and scope and drive development effort. You possess good communication skills and are enthusiastic about AI technologies and strongly motivated to enable our customers with the best feature-rich efficient solutions. THE ROLE: We are looking for an experienced ML/AI software engineer with deep expertise in GPU kernel development and building high‑performance primitives for training and inference. You’ll design, implement, and optimize custom Triton kernels for core ML workloads integrate them into our frameworks and services, and drive end‑to‑end performance on AMD ROCm platform across AMD Radeon and Ryzen product families. KEY RESPONSIBILITIES: Design, implement, and maintain Triton GPU kernels for state‑of‑the‑art ML workloads, with a focus on fusion, tiling, vectorization, and memory‑hierarchy optimization for AMD RDNA GPU architectures. Analyze performance using profiling tools; identify bottlenecks in memory hierarchy, tensor core/matrix core utilization, warp/wavefront scheduling, and synchronization Optimize kernels across problem shapes, batch sizes, and sequence lengths; implement autotuning strategies and performance heuristics. Integrate custom kernels with PyTorch (torch.compile, torch._inductor), JAX, or other frameworks; develop Python/C++ bindings and align with runtime/graph compilers. Track and contribute to upstream Triton and related compiler ecosystems; propose enhancements aligned with our workloads. PREFERRED EXPERIENCE: Triton GPU kernel development and optimization experience CUDA or HIP kernel development experience (porting, performance tuning, and feature enablement) Experience optimizing kernels on AMD GPUs and familiarity with the ROCm software stack (HIP runtime, rocBLAS, MIOpen, rocWMMA, RCCL, etc.) is a plus Familiarity with compiler internals and IRs (LLVM, MLIR) and codegen for GPUs Strong background in linear algebra, convolution algorithms, attention mechanisms, or other core ML primitives Experience integrating kernels with PyTorch, JAX, or TensorFlow, including custom ops/extensions Knowledge of quantization (INT8/FP8), mixed precision, custom dtypes, and numerics (stability, error analysis). Experience with LLM, diffusion and MoE workloads: FlashAttention‑style kernels, paged attention, grouped‑query attention, rotary embeddings, fused MLPs. Contributions to open-source projects in Triton, ROCm, or related GPU/ML ecosystems are a plus ACADEMIC CREDENTIALS: Bachelor’s, Master, or PhD in Computer Science, Electrical Engineering or relevant fields. LOCATION: Belgrade or Nis (Serbia) #LI-CM1 #LI-HYBRID