- 地点
- 科隆，德国
- 工作方式
- 混合办公
- 类型
- 实习
职责
- 设计并实现基于 MLIR 的下一代编译器的新组件
- 探索新型优化技术
- 与客户反复沟通，设计出最佳的编译器技术和最佳的用户体验
- 与不断壮大的开源社区合作，围绕 Roofline 构建强大的生态系统
- 可供选择的论文选题：MLIR 解释器；利用 RISC-V 的向量指令优化性能；Blockbuster（新的平铺策略）
必备要求
- 就读于欧洲顶尖大学的学生（与亚琛工业大学、慕尼黑大学和洛桑联邦理工学院有合作关系，有助于这些大学的学生找到导师）
- 已注册计算机科学/电气工程或相关领域的硕士学位课程
- 具备 C++ 和 Python 的知识和实践经验
- 具备扎实的软件工程原理、数据结构、设计模式和算法知识
- 渴望与不断壮大的开源贡献者社区合作
- 具备解决复杂问题的雄心和好奇心，拥有强烈的责任感，能够主动承担任务，推动解决方案的进展，并拥有协作精神
优先条件
- 有使用 MLIR 或 LLVM 的经验
- 具备人工智能算法知识，有使用人工智能加速器或编译器到相关硬件的经验
- 掌握数据运动网络、数据流编程模型和优化人工智能库的相关知识
- 具备 CPU、GPU 或 NPU/加速器硬件架构方面的知识者优先考虑
Responsibilities
- Design and implement new components of a next-generation MLIR-based compiler
- Explore novel optimization techniques
- Iterate with customers to design the best compiler technology and user experience
- Collaborate with the growing open-source community to build a strong ecosystem around Roofline
- Available thesis topics: MLIR interpreter; optimizing performance with RISC-V vector instructions; Blockbuster (new tiling strategy)
Must Have
- Enrolled at a top European university (partnerships with RWTH Aachen, LMU Munich, and EPFL help students from these universities find supervisors)
- Registered in a Master's program in Computer Science/Electrical Engineering or a related field
- Knowledge and hands-on experience with C++ and Python
- Solid knowledge of software engineering principles, data structures, design patterns, and algorithms
- Eager to collaborate with a growing open-source contributor community
- Ambition and curiosity to solve complex problems, strong sense of responsibility, ability to take initiative, drive solutions forward, and collaborative spirit
Preferred
- Experience with MLIR or LLVM
- Knowledge of AI algorithms and experience with AI accelerators or compilers targeting relevant hardware
- Knowledge of data movement networks, dataflow programming models, and optimized AI libraries
- Knowledge of CPU, GPU, or NPU/accelerator hardware architectures
公司简介
Roofline 正在构建一个软件开发工具包，用于在边缘端的高性能硬件上运行任何模型。在 RooflineAI，我们助力您实现梦寐以求的边缘人工智能产品。我们的使命很简单：将芯片的强大功能带给创新者，并让他们能够更轻松地在顶尖硬件上使用尖端人工智能模型。