Refine Reset All
Sort by
Job Type
Employer/Recruiter
Date Posted
Location
Job Type
Employer/Recruiter
All Filters

Reflection Jobs In Beijing (Peking) - 4 Job Positions Available

1 – 4 of 4 jobs
懂车帝 Dongchedi jobs

大模型算法工程师(LLM)-校招【27届】 北京 AI车青年 职位 ID:A214507 职位描述 1. 负责大语言模型(LLM)及智能体(Agent)在汽车垂类场景的端到端落地,覆盖AI销售、智能客服等公司级核心产品功能;2. 构建基于 智能体协作(Tools/Skills/Handoffs) 的智能体系统,实现多步骤任务规划能力;3. 探索并应用 思维链(Chain-of-Thought, CoT)、自我反思(Reflection)等机制,提升模型在复杂购车咨询、对比分析、金融贷款等高阶推理任务中的表现;4. 参与LLM强化学习优化链路建设,包括偏好数据构建、可验证奖励强化学习(RLVR);5. 与产品、工程、业务团队紧密协作,推动算法模型从实验到上线的全链路工程化落地,并建立科学的效果评估与迭代机制; 职位要求 1. 计算机、人工智能、自动化、数学或相关专业,硕士及以上学历;2. 扎实的机器学习理论基础,深入理解Transformer架构及主流大模型(如LLaMA、Qwen、DeepSeek、GPT系列)的训练与推理机制;3. 具备完整的LLM应用落地经验,熟悉 Prompt Engineering、SFT、DPO 等技术路径,并有实际项目验证效果;4. 对智能体(Agent)技术栈有实践积累,包括但不限于:基于CoT的任务分解与推理、多工具调用(Function Calling)与执行反馈闭环、自主规划(Planning)与状态记忆(Memory)机制、基于用户行为或模拟反馈的策略优化方法5. 熟练掌握Python,具备良好的工程实现与系统设计能力,了解服务端部署、模型监控与AB测试流程;6. 精通PyTorch/TensorFlow等框架,了解分布式训练、推理加速(vLLM/TensorRT-LLM等)、模型量化等优化技术;7. 具备优秀的逻辑思维、业务抽象能力与跨团队协作意识,能将复杂用户需求转化为可量化的技术方案;加分项(优先考虑)1. 有垂直领域(汽车、电商、客服等)智能体或强化学习落地经验;2. 在ICLR、NeurIPS、ICML、ACL、EMNLP等顶会发表过LLM、Agent或RL相关论文;3. 熟悉多智能体协作(handoffs architecture)、可验证奖励强化学习(RLVR);4. 参与过开源Agent框架(如LangChain、LlamaIndex、AutoGen)的二次开发或优化;5. 具备大模型评估体系设计经验,尤其在事实一致性、工具调用准确率、任务完成率等维度有方法论沉淀。 投递...

Premium Full-time
懂车帝 Dongchedi  28 days ago
Cummins jobs

Job Summary: Works in a variety of finance roles, during a two-year developmental program. Based on assigned location, responsibilities will vary, including the preparation of analyses, forecasts and reports, support of the annual budget process, and

Cummins  17 days ago
Cummins jobs

Job Summary: Leads the Talent Acquisition strategy for a moderately sizeable geographic, across multiple sites or countries, with some stakeholder or model complexity. This role focuses on technologically driven processes, operational excellence and delivering best-in-class and

Cummins  13 days ago
VAST jobs

资深 Agent 工程师(Tripo Studio · Canvas Agent) Beijing Experienced Full-time Responsibilities 【我们在做什么】Tripo 做 3D 生成:把文本、图片和交互变成可交付的 3D 资产。https://tripo3d.ai本岗位在 Agent 团队,负责 Studio 中的 Canvas Agent:把自然语言、参考图和画布操作,编排成可执行的创作流程。工作重点是 Agent 架构、评测,以及效果、成本、时延之间的取舍。【你将负责】1. 端到端创作链路:基于自然语言、参考图和画布交互,规划并执行「生成 → 检查 → 分割 / 局部编辑 → 贴图 → 重拓扑

VAST  7 days ago

Subscribe for job alerts and resources to make your job search easier!

Confirmation email sent to

Check your email and click on the link to start receiving your job alerts

Receive the latest job openings for:

reflection jobs in beijing

Confirmation email sent to

Check your email and click on the link to start receiving your job alerts

All Filters Apply
Sort by
Job Type
Employer/Recruiter