AI算法工程师（Agent Model Post-Training）-即梦
Core
Build post-training capabilities for a creative agent model to enhance its performance in creative understanding, task planning, tool invocation, and multi-turn interactions.
Role type
Senior IC machine-learning engineer (agentic AI)
Builds
Creative agent models for multi-modal generation and interaction
Domain
Generative AI / Agentic Systems / Multi-modal
Deliverable
production ML models
Required skills
Large model post-training (SFT, RLHF, DPO, PPO, GRPO, RLAIF), Agentic RL, Tool Use, Multi-turn Function Call, Multi-modal model training, PyTorch, Experimental design, Data analysis
Preferred skills
None stated
Technologies
PyTorch
Responsibilities
Construct SFT data, multi-turn Function Call data, tool trajectory data, and creative task preference data; Explore Agentic RL and reward systems to improve long-chain task success rates; Collaborate with product and evaluation teams to establish continuous iteration loops for training and assessment.
