腾讯云DataBuddy-大模型算法专家
Core
Build evaluation systems and post-training pipelines for DataBuddy, an enterprise-grade large model for big data tasks, focusing on task understanding, data Q&A, tool use, and result reliability.
Role type
Senior IC large model algorithm expert (post-training & evaluation)
Builds
Evaluation frameworks, data flywheels, and post-training pipelines (SFT, RLHF/RLAIF) for DataBuddy
Domain
Enterprise big data / Large Language Models
Deliverable
production ML models
Required skills
Large model post-training (SFT, RLHF/RLAIF), evaluation system design, data construction, prompt engineering, tool use evaluation, Python, PyTorch, Transformers, vLLM, data analysis, problem attribution
Preferred skills
Experience with BI, data governance, data development, and data analysis scenarios
Technologies
PyTorch, Transformers, vLLM, Python
Responsibilities
Build evaluation systems covering task understanding, data Q&A, tool calling, and result reliability; Construct data flywheels from user feedback and logs; Apply post-training methods to improve model stability and accuracy; Design automated evaluation and benchmarking mechanisms; Collaborate with product and engineering teams to close the loop on model optimization.
Seniority
Senior, hands-on IC