推荐大模型系统研发工程师 - Data AML
Core
Building and scaling recommendation foundation models and high-performance distributed training systems for TikTok, Toutiao, and external clients.
Role type
Senior IC machine-learning engineer (recommendation systems & LLMs)
Builds
Recommendation-specific generative models, ND-Parallel libraries, and cross-business foundation models
Domain
Recommendation systems, Large Language Models (LLM), Distributed Computing, High-Performance Computing
Deliverable
production ML models
Required skills
C++, Python, CUDA, Triton, TensorFlow, PyTorch, JAX, Distributed Systems, Parallel Strategies (Data/Model/Expert/Sequence/Pipeline), Deep Learning Frameworks, Recommendation Algorithms (DeepFM, DLRM, DIN), Transformer, BERT, MoE, CLIP
Preferred skills
100B+ scale LLM pretraining/fine-tuning/RL, Multi-modal representation learning (CLIP, BLIP, VLM), GPU architecture & NCCL/RDMA optimization, Framework codebase deep dives
Responsibilities
Scaling recommendation model sequence length, parameter scale, and training volume; Building training/inference systems for generative models on billion-user data; Developing ND-Parallel libraries for large-scale cluster stability; Constructing cross-business recommendation foundation models via distillation; Integrating multi-modal signals (text, image, video) and LLMs (CPT/SFT/RL) into recommendation systems