📋 Job Overview

The role involves developing and optimizing a large model training framework for Tencent's Hunyuan model, supporting efficient and stable training at scales exceeding 10,000 GPUs for single tasks. Responsibilities include designing NLP and multimodal model architectures, verifying training efficiency and effectiveness in collaboration with business teams, and accelerating training performance for applications like text-to-image, text-to-video, and text-to-3D generation. The position also focuses on optimizing low-precision training and large-window training performance for broader business adoption.