We own the platform blueprint for our ML infrastructure: designing systems that integrate with a data mesh of domain-owned data products, leverage Qualcomm Cloud AI 100 and NVIDIA GPU clusters for training at petabyte scale and produce optimised model artefacts ready for deployment to vehicle hardware. * Also, you design the data-format backbone, setting standards for data flows, ingestion, cataloguing, transcoding, and partitioning at PB scale, integrated with dataset management tooling. * Further, you drive cost governance at PB scale, including accelerator spot strategies, S3 tiering, cross-AZ traffic reduction, and Kubernetes cluster right-sizing. * Proven track record designing systems for PB-scale data and hundreds of concurrent training jobs as well as understanding of large vision models and the challenges of compressing them for automotive-grade SoCs.
mehr