代码智能体大模型架构实习生（J101404）
Core
Building training infrastructure and data pipelines for Coding Agents, focusing on trajectory synthesis, cleaning, and large-scale model training/inference optimization.
Role type
Intern, Machine Learning Infrastructure Engineer (Coding Agents)
Builds
Coding Agent training environments, trajectory synthesis datasets, and inference services
Domain
Artificial Intelligence / Large Language Models / Software Engineering
Deliverable
production ML models
Required skills
Large language model training, distributed training frameworks, trajectory synthesis, data cleaning, performance analysis
Preferred skills
Coding agent harnesses, SFT/DPO/GRPO, CUDA optimization, inference engines
Technologies
Kubernetes, Transformers, DeepSpeed, FSDP, Megatron-LM, verl, vLLM, SGLang, JSONL
Responsibilities
Build Coding Agent sandbox training environments, synthesize and clean large-scale trajectory data, adapt training frameworks, analyze training and inference performance
Seniority
Intern
