Research Engineer/Scientist(all levels), Efficient Models
Core
Designing and implementing efficient models for large-scale generative AI, focusing on model distillation and compression to enable scalable training and deployment.
Role type
Research Engineer/Scientist (Efficient Models)
Builds
Scalable generative modeling approaches (diffusion, autoregressive) and efficient inference frameworks for TikTok products.
Domain
Generative AI, Computer Vision, Multimodal Understanding
Deliverable
production ML models
Required skills
Model distillation, model compression, quantization, step distillation, cfg distillation, generative AI training, LLM training, PyTorch, JAX
Preferred skills
Diffusion models, autoregressive models, MoE architectures, window attention, reinforcement learning training methods (RLHF, DPO, GRPO), image/video generation and editing
Technologies
PyTorch, JAX
Responsibilities
Develop efficient algorithms and architectures for large-scale generative and multimodal models; Advance scalable generative modeling approaches with a focus on acceleration and efficiency.
Seniority
All levels (Entry to Senior)
