多模态算法工程师（模型优化方向）/ Multimodal Algorithm Engineer (Model Optimization) at Advanced Micro Devices, Inc

Pudong, Shanghai, China -

Full Time

Start Date

Immediate

Expiry Date

25 Aug, 26

Salary

0.0

Posted On

27 May, 26

Experience

2 year(s) or above

Remote Job

Yes

Telecommute

Yes

Sponsor Visa

Skills

Multimodal Foundation Models, Vision-Language Models, Model Quantization, PyTorch, Model Compression, Knowledge Distillation, Embodied AI, Robotics Decision Making, Operator Fusion, KV Cache Optimization, AMD ROCm, CUDA, HIP, Simulation Platforms, Software Engineering, Heterogeneous Computing

Industry

Semiconductor Manufacturing

Description

WHAT YOU DO AT AMD CHANGES EVERYTHING At AMD, our mission is to build great products that accelerate next-generation computing experiences—from AI and data centers, to PCs, gaming and embedded systems. Grounded in a culture of innovation and collaboration, we believe real progress comes from bold ideas, human ingenuity and a shared passion to create something extraordinary. When you join AMD, you’ll discover the real differentiator is our culture. We push the limits of innovation to solve the world’s most important challenges—striving for execution excellence, while being direct, humble, collaborative, and inclusive of diverse perspectives. Join us as we shape the future of AI and beyond. Together, we advance your career. THE ROLE: We are a core algorithm team at AMD, dedicated to end-to-end AI workload optimization on AMD platforms. We are seeking talented engineers specializing in multimodal foundation models, with a focus on Vision-Language Models (VLMs), Vision-Language-Action Models (VLAs), and World Action Models (WAMs). In this role, you will drive model training, compression, quantization, inference optimization, and efficient deployment—enabling next-generation embodied AI and multimodal agents to achieve peak performance on AMD hardware platforms. KEY RESPONSIBILITIES: Optimize training strategies, fine-tuning, and alignment for multimodal models (VLM / VLA / WAM) on AMD platforms. Enhance action prediction, world state modeling, and long-horizon planning capabilities of WAM/VLA models for embodied intelligence scenarios (e.g., robotics, simulation-based interaction). Design and implement model optimization techniques including quantization (PTQ/QAT), pruning, knowledge distillation, operator fusion, and KV cache optimization to improve inference latency, throughput, and energy efficiency. Collaborate closely with compiler, driver, and system software teams to deeply integrate models into AMD’s software stack. Stay at the forefront of research in World Models, action generation, and multimodal agents—and explore novel architectures for AMD’s heterogeneous compute platforms. QUALIFICATIONS: Master’s or PhD in Computer Science, Artificial Intelligence, Robotics, Electrical Engineering, or a related field. Hands-on experience with VLMs, VLAs, or WAMs (World Action Models)—especially in robotics decision-making, simulated environment training, or action sequence generation—is highly preferred. Proficiency in PyTorch; familiarity with multimodal and embodied AI frameworks. Familiarity with simulation platforms such as Isaac Gym, LIBERO, MuJoCo, or RoboTwin. Strong software engineering skills and ability to deliver full-cycle solutions—from research prototyping to production deployment. PREFERRED QUALIFICATIONS: Contributions to open-source projects in multimodal agents, world models, or robotics (e.g., OpenVLA, DROID, ACT). Publication record in top-tier conferences (e.g., CVPR, ICRA, CoRL, NeurIPS, ICLR) in multimodal learning or embodied AI is a strong advantage. Strong background in model optimization: quantization, sparsity, kernel fusion, dynamic batching, etc. Experience with AMD ROCm ecosystem or heterogeneous computing performance tuning. Understanding of GPU/accelerator architecture; experience with CUDA or HIP is a plus. WHAT WE OFFER: Access to cutting-edge AMD compute resources. Unique opportunity to shape full-stack co-design across algorithms, compilers, and hardware. A collaborative, globally distributed team of world-class AI systems and robotics researchers. #LI-JW2 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. AMD may use Artificial Intelligence to help screen, assess or select applicants for this position. AMD’s “Responsible AI Policy” is available here. This posting is for an existing vacancy.

Responsibilities

Optimize training, fine-tuning, and alignment for multimodal models (VLM, VLA, WAM) on AMD platforms. Implement model optimization techniques like quantization and pruning to improve inference latency and energy efficiency for embodied AI.