Machine Learning Engineer II, AWS Just-Walk-Out Science Team Join to apply for the Machine Learning Engineer II, AWS Just-Walk-Out Science Team role at Amazon Web Services (AWS) Role Overview As part of the AWS Solutions organization, we aim to provide innovative business applications that are used by millions worldwide to manage daily operations. Our goal is to accelerate our customers’ businesses through intuitive and differentiated technological solutions that address enduring business challenges. We combine vision with curiosity and Amazon’s real-world experience to build turnkey solutions that are easy to adopt and use, especially where customers prefer buying over building. The Team Just Walk Out (JWO) is a revolutionary retail concept with no lines or checkout—customers just grab and go using the Amazon Go app. Our checkout-free shopping experience relies on Just Walk Out Technology, which uses computer vision, sensor fusion, deep learning, and foundation models to detect product interactions and manage virtual carts. Our team is pushing the boundaries of computer vision, deep learning, real-time systems, and hardware design, working in a startup-like, collaborative environment to develop autonomous AI agents that understand scenes, customer behavior, and adapt dynamically. Key Responsibilities Collaborate with Applied Scientists to integrate cutting-edge model architectures into training pipelines and auto-labeling processes. Process large datasets, scale machine learning models, optimize GPU utilization, memory, and training workflows (kernel fusion, mixed-precision training, gradient accumulation, offloading optimizer states, parallelization). Design and maintain distributed training systems supporting multi-modal foundation models for autonomous retailing, optimizing GPU usage for large datasets. Develop monitoring and debugging tools to ensure training workflow reliability and performance on GPU clusters; maintain large-scale auto-labeling pipelines. Work with engineers and scientists to prototype new technologies, evaluate feasibility, and solve complex problems. Day in the Life As a Machine Learning Engineer on the JWO team, you will lead development of algorithms and models to advance training techniques using hardware like NVIDIA GPUs. Your work will directly impact our products and customer experience, leveraging Amazon’s diverse data sources and large-scale computing resources. You will influence our strategic direction, system architecture, and best practices, working in an Agile environment to deliver high-quality software. Basic Qualifications 3+ years of professional software development experience, including coding standards, code reviews, source control, build, testing, and operations. 2+ years of experience in system design or architecture, including reliability and scaling. Proficiency in Python or related languages. Hands-on experience with PyTorch, deep learning frameworks like MMEngine or Megatron-LM, and large-scale deep learning operations. Knowledge of visual-language models, multi-modal AI, pre-training/post-training techniques, and performance profiling tools. Preferred Qualifications Master's or PhD in computer science or related field. Over a year of experience in deploying or optimizing ML models, with strong skills in scalable GPU training frameworks and familiarity with HuggingFace Transformers. Experience with large-scale multimodal LLM and generative models, contributions to open-source frameworks, or research publications. Expertise in GPU optimization techniques, mixed precision training, model parallelism, and efficient data preprocessing pipelines. Proven track record in video understanding and multi-modal learning, including designing scalable architectures and evaluation at scale. Amazon is an equal opportunity employer. We consider qualified applicants with arrest and conviction records, and support accommodations for applicants with disabilities. Compensation varies by location and experience, with additional benefits. For more details, visit our careers page. This position remains open until filled. Applicants should apply via our career site.

#J-18808-Ljbffr