AI back-end Engineer
IBM
IBM
Join IBM Infrastructure & Technology as an AI Back-End Engineer, where you'll contribute to the systems powering the global infrastructure. We focus on high-resiliency mainframes, hybrid cloud platforms, networking, automation, and site reliability. This role involves tackling complex challenges with curiosity and collaboration, working with diverse technologies and global colleagues to deliver future-ready solutions.
As an AI Back-End Developer specializing in IBM Z, you will operate within an agile, collaborative setting. You'll design, code, and deliver innovative solutions to client problems and requirements, utilizing IBM's Design Thinking to ensure exceptional user experience, quality, resiliency, performance, security, and stability.
Key responsibilities include designing, coding, debugging, testing, and delivering creative solutions for client needs in AI-related domains, applying DevOps methodologies. You will collaborate closely within an agile team to understand requirements and deliver high-quality outcomes. Employing IBM's Design Thinking is crucial for creating user-centric solutions that excel in quality, resiliency, performance, security, and stability.
This role involves working on cutting-edge projects in machine learning, deep learning, and other AI areas to foster innovation and drive solution delivery. Ensuring that all solutions meet stringent standards for quality, resiliency, performance, security, and stability is paramount.
A minimum of 5 to 10 years of industrial experience is required, specifically with Unix/Linux-based products developed using Python, C, or C++. Significant experience in AI development, including designing, coding, debugging, and testing PyTorch Core & Internals, is essential. Contributions to and extensions of PyTorch internals and back-end architecture, such as TorchInductor and TorchDynamo, are highly valued.
Candidates must possess deep hands-on experience with PyTorch FSDP and the ability to debug low-level PyTorch runtime issues. Optimization of model execution paths for training and inference, along with experience in proprietary or custom AI accelerators, is expected. Advanced C++ and Python skills, experience with GPU programming or accelerator SDKs, and strong debugging abilities are necessary. Familiarity with DevOps methodologies and a deep understanding of AI runtimes, memory hierarchies, parallel execution models, PyTorch distributed runtime, parameter sharding, and memory management are also critical.
IBM
Technology