算法优化工程师（多模态生成） - Seed Model
Core
Develop algorithm acceleration techniques for text-to-image and text-to-video generative models to achieve order-of-magnitude efficiency improvements.
Role type
Senior IC algorithm optimization engineer (multimodal generation)
Builds
Optimized inference engines for diffusion and autoregressive models supporting Doubao, Jimeng, and TRAE applications
Domain
Generative AI, Multimodal Learning, Model Compression
Deliverable
production ML models
Required skills
Diffusion model architecture, Autoregressive model architecture, Model quantization, Knowledge distillation, MoE architecture, Sparse attention mechanisms, DPO alignment algorithms, Inference bottleneck analysis, Algorithmic optimization methodologies
Preferred skills
Kaggle competition awards, Publications in ICML/NeurIPS/CVPR/ECCV/ICCV, Core contributions to open-source frameworks like HuggingFace Diffusers
Technologies
Diffusion models, Autoregressive models, MoE, Sparse networks, Low-bit quantization
Responsibilities
Research and develop algorithm acceleration for multimodal generation models; Lead innovative algorithm research to improve inference efficiency; Collaborate with upstream/downstream teams to analyze performance bottlenecks and optimize model reasoning efficiency
Seniority
Senior, hands-on IC
