AMDTriton Compiler Development Engineer

THE ROLE: Triton is a language and compiler for writing highly efficient custom deep learning primitives. It's widely adopted in open AI software stack projects like PyTorch, vLLM, SGLang, and many others. AMD GPU is an official backend in Triton and we are fully committed to it. If you are interested in making GPUs running fast via developing the Triton compiler and kernels, please come join us!

THE ROLE: MTS Software development engineer on teams building and optimizing Deep Learning applications and AI frameworks for AMD GPU compute platforms.  Work as part of an AMD development team and open-source community to analyze, develop, test and deploy improvements to make AMD the best platform for machine learning applications. THE PERSON: Strong technical and analytical skills in C++ development in a Linux environment. Ability to work as part of a team, while also being able to work independently, define goals and scope and lead your own development effort. KEY RESPONSIBILITIES: Optimize Deep Learning Frameworks: In depth experience in enhance and optimize frameworks like TensorFlow and PyTorch for AMD GPUs in open-source repositories. Develop GPU Kernels: Create and optimize GPU kernels to maximize performance for specific AI operations. Develop & Optimize Models: Design and optimize deep learning models specifically for AMD GPU performance. Collaborate with GPU Library Teams: Work tightly with internal teams to analyze and improve training and inference performance on AMD GPUs. Collaborate with Open-Source Maintainers: Engage with framework maintainers to ensure code changes are aligned with requirements and integrated upstream. Work in Distributed Computing Environments: Optimize deep learning performance on both scale-up (multi-GPU) and scale-out (multi-node) systems. Utilize Cutting-Edge Compiler Tech: Leverage advanced compiler technologies to improve deep learning performance. Optimize Deep Learning Pipeline: Enhance the full pipeline, including integrating graph compilers. Software Engineering Best Practices: Apply sound engineering principles to ensure robust, maintainable solutions.

An exciting internship opportunity to make an immediate contribution to AMD's next generation of technology innovations awaits you! We have a multifaceted, high-energy work environment filled with a diverse group of employees, and we provide outstanding opportunities for developing your career. During your internship, our programs provide the opportunity to collaborate with AMD leaders, receive one-on-one mentorship, attend amazing networking events, and much more. Being part of AMD means receiving hands-on experience that will give you a competitive edge. Together We Advance your career! JOB DETAILS: Location: Beijing,China Onsite/Hybrid: at least 3 days a week, either in a hybrid or onsite or remote work structure throughout the duration of the co-op/intern term. Duration: at least 6 months WHAT YOU WILL BE DOING: We are seeking highly motivated AI Compiler Software Engineering intern/co-op to join our team. In this role – We will involve you in extending Triton’s compiler infrastructure to support new AI workloads and hardware targets. We will assign you tasks to implement and optimize GPU kernels using Triton’s Python-based DSL. We will train you to analyze kernel performance using profiling tools and help you identify bottlenecks and optimization opportunities. We will understand how modern compilers translate high-level abstractions into efficient machine code.

Position Overview We are seeking a highly experienced engineer specializing in large language model (LLM) inference performance optimization. You will be a core member of our team, responsible for building and optimizing the LLM inference performance with high-throughput, low-latency on AMD Instinct GPUs. If you are passionate about pushing performance boundaries and have deep, hands-on expertise with cutting-edge technologies like vLLM or SGLang, we invite you to join us. Key Responsibilities 1. Core System Optimization: Lead the development, tuning, and customization of LLM performance optimization on AMD GPUs, leveraging and extending frameworks like vLLM or SGLang to address performance bottlenecks in production environments. 2. Performance Analysis & Tuning: Conduct end-to-end performance profiling using specialized tools. Perform deep optimization of compute-bound operators (e.g., Attention), memory I/O, and communication to significantly increase throughput and reduce latency. 3. Model Architecture Adaptation: Demonstrate expertise in mainstream LLM architectures (e.g., DeepSeek, Qwen, Llama, ChatGLM) and optimize inference for their specific characteristics (e.g., RoPE, SWA, MoE, GQA). 4. Algorithm & Principle Application: Leverage your deep understanding of core algorithms (Transformer, Attention, MoE) to implement advanced optimization techniques such as PagedAttention, FlashAttention, continuous batching, quantization, and model compression. 5. Technology Foresight & Implementation: Research and prototype state-of-the-art optimization techniques (e.g., Speculative Decoding, Weight-Only Quantization) and drive their adoption into production systems. Qualifications: Mandatory