混元大语言模型后训练算法工程师-模型合版方向（北京/深圳）
Core
Develop post-training strategies (SFT, RL) and model merging algorithms to enhance generalization, capability boundaries, and performance limits of large language models.
Role type
Senior IC large language model post-training algorithm engineer (model merging)
Builds
Post-trained large language models with enhanced reasoning, agentic capabilities, and efficiency
Domain
Artificial Intelligence / Large Language Models
Deliverable
production ML models
Required skills
SFT, RL, reward modeling, model merging, PyTorch, Megatron, vLLM, data analysis, algorithm strategy research
Preferred skills
Reasoning/Agentic capability enhancement, Scaling laws, RL/OPD convergence research, technical report/paper writing
Technologies
PyTorch, Megatron, vLLM
Responsibilities
Research and develop post-training strategies and production paradigms for large models; Explore frontier techniques including reasoning, agentic capabilities, and scaling; Convert research findings into actual model capabilities or production efficiency improvements.
Seniority
Senior, hands-on IC
