【AI Infra】模型闭环工程师(J12898)

免费注册后解锁"你的匹配分 vs 高分画像"与个人技能缺口。

1、做什么：
打通"数据→训练→推理→仿真→评测→回流"全链路闭环，探索 AI Agent 驱动的自动化训练与数据自进化能力，支撑具身模型天级迭代。

2、核心职责：
- 训练-推理-仿真-评测全链路打通，设计自动化管线支撑天级迭代
- 仿真评测体系建设（Sim2Real 验证、开环/闭环双轨评测、结果自动回流）
- AI Agent 驱动的自动化训练探索（模型短板识别 → 数据采集/合成 → 自动训练 → 评测 → 部署决策）
- 模型全生命周期工具链建设（版本管理、实验追踪、效果可视化）
- 计算机/AI 等相关专业，硕士及以上
- 熟悉 PyTorch/TensorFlow，有分布式训练（DDP/FSDP/DeepSpeed）基础
- 了解模型服务化部署（Triton/TensorRT/vLLM）
- 了解仿真环境（Isaac Sim/Mujoco/PyBullet）或评测体系建设
- 对 AI Agent、AutoML 有研究或实践兴趣

加分项：
- 具身智能/VLA 模型训练与部署经验
- 强化学习（PPO/SAC/RLHF）训练管线经验
- 开源 ML 平台（MLflow/WandB/Kubeflow）深度使用

1. What to do:
Connect the full closed loop of "data → training → inference → simulation → evaluation → feedback" and explore AI Agent-driven automated training and data self-evolution capabilities to support day-level iteration of embodied models.

2. Core responsibilities:
- Connect the full chain of training-inference-simulation-evaluation, and design automated pipelines to support day-level iteration
- Build a simulation evaluation system (Sim2Real validation, open-loop/closed-loop dual-track evaluation, automatic feedback of results)
- Explore AI Agent-driven automated training (model weakness identification → data collection/synthesis → automated training → evaluation → deployment decisions)
- Build a toolchain for the full model lifecycle (version management, experiment tracking, performance visualization)
- Major in Computer Science/AI or related fields, master's degree or above
- Familiar with PyTorch/TensorFlow, with a foundation in distributed training (DDP/FSDP/DeepSpeed)
- Understanding of model service deployment (Triton/TensorRT/vLLM)
- Understanding of simulation environments (Isaac Sim/Mujoco/PyBullet) or evaluation system construction
- Research or practical interest in AI Agents and AutoML

Bonus points:
- Experience in embodied intelligence/VLA model training and deployment
- Experience in reinforcement learning (PPO/SAC/RLHF) training pipelines
- Deep use of open-source ML platforms (MLflow/WandB/Kubeflow)

⚑ 岗位信息收录自公开招聘页（国内官方招聘渠道），以雇主官网为准