AMD is seeking a highly motivated PMTS AI/ML Compiler Engineer to join its AI Software organization. This role focuses on driving compiler and performance optimizations for AI training and inference workloads on AMD GPU platforms. The ideal candidate will work at the intersection of compiler technology, machine learning frameworks, and performance engineering. The position requires collaboration across various teams to enhance the ROCm software ecosystem and ONNX Runtime execution stack.

• Strong proficiency in C++ and object-oriented software development.
• Hands-on experience with GPU programming technologies such as HIP, CUDA, Triton, or equivalent parallel programming frameworks.
• Experience with compiler technologies including MLIR, LLVM, or related compiler infrastructures.
• Strong understanding of machine learning frameworks, model execution pipelines, and AI workload optimization.
• Experience profiling and optimizing performance-critical software systems.
• Proficiency with Linux development environments, source control systems (Git), debugging tools, and performance profilers.
• Strong analytical, communication, and problem-solving skills.
• Bachelor’s or Master's degree in Computer Science, Computer Engineering, Electrical Engineering, or equivalent.

• Work with AMD’s architecture specialists to improve future products.
• Design, develop, and optimize compiler transformations and operator lowerings within MLIR, LLVM, and the ROCm software stack.
• Develop scalable solutions that improve portability and performance across AMD GPU architectures.
• Collaborate with compiler and runtime teams to develop advanced optimization strategies for AI workloads.
• Implement, validate, and maintain ONNX operators and microbenchmark test suites.
• Lead performance optimization efforts for AI training, fine-tuning, reinforcement learning (RL), and inference workloads on AMD GPUs.
• Design and implement graph-level, kernel-level, and runtime-level optimizations that deliver measurable performance gains.
• Analyze workload execution characteristics and identify opportunities for improvements in compute efficiency, memory utilization, and communication overhead.
• Drive performance characterization and benchmarking across pre-silicon and post-silicon environments.

• AMD provides a competitive 'Total Rewards' package that focuses on financial growth, health, and work-life balance.