Research Engineer/Scientist (all levels), Efficient Models

Work on Vision-Applied Research for ByteDance’s Generative AI and CV/multimodal products. Design and implement efficient large-scale generative model methods focused on large model distillation, compression, scalable training, optimization, and hardware-efficient inference. Build distillation frameworks, accelerate diffusion and autoregressive models, and apply efficiency techniques such as quantization, step/cfg distillation, and other model-architecture approaches to enable deployment at scale.