Applied Research Scientist - AI Models & Agents

优缺点分析

角色解读

• Progress to senior research scientist leading independent research directions.
• Transition into technical leadership roles (e.g., principal scientist, research manager).
• Opportunity to shape AMD's AI strategy and influence next-gen hardware design.
• Research and improve state-of-the-art generative AI models, including LLMs, multimodal models, and diffusion models.
• Optimize training and inference pipelines for AMD hardware accelerators.
• Develop autonomous AI agents for tasks like code generation and kernel optimization.
• Publish research findings at top conferences and contribute to open-source communities.
• Deep expertise in Python and deep learning frameworks (PyTorch/TensorFlow).
• Solid understanding of transformer architectures, state space models, and attention mechanisms.
• Experience with kernel writing in CUDA, HIP, or Triton for GPU acceleration.
• Familiarity with distributed training, agentic AI frameworks, and hardware-efficient inference.

申请策略

• Tailor your application to demonstrate alignment with AMD's research focus on hardware-software co-design.
• Prepare to discuss your research vision and how it complements AMD's AI roadmap.
• Emphasize publications at top conferences (NeurIPS, ICML, ICLR).
• Showcase open-source contributions and projects on GitHub.
• Highlight hands-on experience with LLM training, distributed systems, and kernel optimization.
• Detail any prior work on AI agents or hardware-aware model design.
• Deepen knowledge of AI agent frameworks (e.g., LangChain, AutoGPT).
• Gain proficiency in GPU programming with HIP or CUDA.

面试指南

• For research questions: Start with problem motivation, then discuss your approach, experiments, and results.
• For optimization questions: Identify bottlenecks, propose specific techniques (e.g., kernel fusion, mixed precision), and consider hardware constraints.
• For system design: Outline high-level architecture, trade-offs, and scalability considerations.
• Describe your most significant research contribution and its impact.
• How would you optimize the training of a large language model on AMD GPUs?
• Explain the advantages and limitations of different attention mechanisms.
• How would you build an autonomous agent for code generation?
• What are the key challenges in deploying large models for inference?