Meituan faced significant computing power and cost challenges while scaling vision AI inference services across massive datasets for lifestyle and e-commerce operations. To overcome these hurdles without relying on expensive discrete GPUs, the company integrated 4th Gen Intel Xeon Scalable processors. Critical to maximizing end-to-end application processing speed, specialized Intel software tools—including oneAPI technologies, the Intel Integrated Performance Primitives, and the Intel Extension for PyTorch—empowered developers to drastically optimize image processing tasks like rotation and resizing. By combining powerful Intel software with advanced hardware accelerators, Meituan boosted inference performance, increased overall resource efficiency threefold, and achieved a 70 percent redu

Join for free to read