Meta - Menlo Park, CA
posted 2 months ago
Meta's AI Training and Inference Infrastructure is experiencing exponential growth to support an increasing array of AI use cases. This growth presents a significant scaling challenge that our engineers must address daily. The primary focus of this role is to build and evolve our network infrastructure, which connects numerous training accelerators, such as GPUs. It is crucial to ensure that the network operates smoothly and meets the stringent performance and availability requirements of RDMA workloads, which demand a loss-less fabric interconnect. To enhance the performance of these systems, we continuously seek opportunities across the entire stack, including network fabric, host networking, communication libraries, and scheduling infrastructure. As an AI/HPC Network Engineer, you will be responsible for designing, developing, testing, and operating networking systems that support large-scale AI training jobs. This involves researching, developing, and deploying various technologies and network topologies to evolve and scale our AI networks effectively. Collaboration is key in this role, as you will work closely with hardware, software, and sourcing teams to develop innovative networking solutions and influence the future of networking and its associated infrastructure. Additionally, you will define and develop optimized network monitoring systems to ensure the reliability and efficiency of our networks. Being on-call will also be part of your responsibilities, allowing you to learn from real-world production challenges and apply those lessons to improve current and future generation products.