PUBLIC ROLE INFORMATION
【27届】VLA算法工程师 - 闭环进化
Key conditions
- Locations
- 北京, 深圳
- Employment
- 未确认
Not disclosed: Team, Track, Salary, Employer application deadline
Sources and available information
Reported open · Last observed 2026-10-18
- Topics
- Unknown
- Company
- 自变量 · Source supported · Coverage unknown
- Work mode
- Unknown
- Source type
- 官方岗位页 · Source supported · Coverage unknown
- Company type
- 创业公司 · Tentative · Coverage unknown
- Hiring track
- Unknown
- Remote scope
- Unknown
- Role
- 算法 · Source supported
- Business / team
- Unknown
- Location
- 北京、深圳 · Source supported
- Employment
- Unknown
- Role AI context
- 直接研发/交付AI · Source supported
- Company AI context
- Unknown
- Recruitment status
- 来源显示在招 · Source supported
Unknown information is retained as such and updated when new source information is available.
- View source ↗ · 2026-10-18
Application path and records
Apply through the official channel. The full external receipt stays in your workspace. When signed in, three minimal event types explicitly recorded through the CLI are reported by default: selection, self-reported application and channel observation. Browsing or saving a role does not report an event; external applications do not appear under My applications.
Role description from source
1、负责探索具身模型架构、训练方法,打造新一代 VLA/WM模型;
2、探索自动化高质量的数采方法,探索有效的接管数采方式,以及接管数据的价值,探索离线和在线强化学习方法;
3、跟踪最新的模型训练和优化技术,复现验证其有效价值,并将其适配到代码框架。
4、参与构造具身模型评测体系,设计评测任务,搭建VLA 模型自动化评测流程;
5、参与VLA 模型相关学术 benchmark 和国内外知名公开榜单的评测,迭代优化模型效果;
职位要求
1、毕业时间介于2026年9月至2027年8月的国内外高校的本科生、硕士生、博士生,计算机科学、机器人控制、人工智能、VLA算等相关专业优先;
2. 掌握常见VLA模型(例如PI系列、OpenVLA、diffusion policy等)原理;
3、熟悉机器人运动控制,有VLA模型训练优化和实际上机部署执行经验;
4、熟练掌握主流深度学习框架,如PyTorch、TensorFlow,具备优秀的系统级编程能力;有LLM、Diffusion、Flow、MoE、VLA等大模型架构的分布式训练经验;
5、精通至少一种分布式训练框架,如Megatron-LM、DeepSpeed、FSDP等;
6、具备良好的逻辑思维、沟通协作能力和学习能力,能够独立解决复杂问题;
7、ACM等程序设计比赛、自然语言处理/多模态相关顶级会议论文或比赛获奖经历者优先考虑;
8、具有技术热情及技术信仰,较强的学习能力和抗压能力。
ABOUT THE COMPANY / TEAM
自变量
Stage unknown