大模型平台策略算法工程师（J97419）
Core
Develop toolchains for post-training large language models (SFT, distillation, RL) on the Qianfan MAAS platform to enhance training stability and efficiency.
Role type
Senior IC machine-learning engineer (LLM post-training platform)
Builds
Post-training toolchains and RLaaS capabilities for internal and external clients
Domain
Artificial Intelligence / Large Language Models
Deliverable
production ML models
Required skills
Python, PyTorch, Transformer architecture, SFT, DPO, RFT, distributed parallel strategies (ZeRO, TP, PP, EP)
Preferred skills
verl, llamafactory, ms-swift, slime, openrlhf frameworks, large-scale cluster optimization (1000+ GPUs)
Responsibilities
Develop toolchains for SFT, distillation, and RL; Implement RLaaS solutions; Optimize platform usability and component reusability; Integrate cutting-edge post-training technologies
Seniority
Senior, hands-on IC
