博世
AI Expert_PS
AI Expert_PS
发布于 1 天前普通员工/个人贡献者
无锡市
高级经验
全职员工
仅现场办公
硕士
机器学习工程
Agentic Ai
Autogen
Crewai
Langgraph
Llm
Mlops
Pytorch
Rag
AI 估算 · 25k–40k
AI专家岗位稀缺，Agentic AI和深度学习技能要求高，博世平台大，薪资竞争力强。
职位详情
关于这个职位
作为博世AI专家，你将负责设计和开发Agentic AI系统，涵盖多智能体架构、LLM应用、计算机视觉等前沿技术
你将推动制造业的智能化和自动化，从概念验证到生产部署全程负责
适合有扎实AI基础并渴望在工业场景应用大模型和智能体的技术人才
最低要求
Qualifications
Master's degree (or above) in Computer Science, Computer Vision, AI/ML, Electrical Engineering, or a closely related field.
+ years of hands-on experience in real AI/ML projects, with at least 1 year in Agentic AI, LLM application development, or equivalent.
Proven track record of delivering AI solutions from prototype to production in an industrial or manufacturing environment.
Strong proactive, self-driven learning mindset with a customer-oriented and CIP (Continuous Improvement) attitude.
Job Relevant Knowledge and Experience
Agentic AI & LLM Engineering**
Hands-on experience with Coding Agents: Claude Code (Anthropic), OpenAI Codex / Codex CLI, GitHub Copilot Workspace, or Cursor.
Experience building and running multi-agent systems with LangGraph, AutoGen, CrewAI, or similar agentic orchestration frameworks.
Proficient in LLM API integration (OpenAI, Anthropic Claude, Azure OpenAI, open-source LLMs via Ollama / vLLM).
Skilled in Retrieval-Augmented Generation (RAG): vector databases (Chroma, Weaviate, Qdrant, pgvector), embedding pipelines, context management.
Familiar with prompt engineering patterns: chain-of-thought, few-shot, tool-calling, structured output, and self-reflection loops.
Understanding of agent safety: output guardrails, hallucination detection, human-in-the-loop design, audit trails.
Familiar with MCP (Model Context Protocol) or comparable agent-tool integration standards.
Computer Vision & Deep Learning**
Expert in Python; proficient in PyTorch and/or TensorFlow/Keras.
Strong in classical image processing (OpenCV): filtering, edge detection, feature extraction, camera calibration.
Deep understanding of CNN architectures: AlexNet, VGG, ResNet, DenseNet, MobileNet, EfficientNet.
Proficient in object detection (YOLO, Faster-RCNN, SSD, DETR) and segmentation (Mask-RCNN, SAM).
Experience with Vision Transformers (ViT, Swin) and multimodal models (CLIP, LLaVA, GPT-4V).
Familiar with data labelling toolchains (Label Studio, CVAT) and data versioning (DVC, MLflow).
Knowledge of model optimisation for edge deployment: quantisation, pruning, ONNX, TensorRT.
MLOps & Engineering Practices**
Proficient in Git and CI/CD pipelines for ML/AI systems.
Experience with containerisation (Docker, Kubernetes) for agent and model serving.
Familiar with cloud/on-prem ML platforms (Azure ML, AWS SageMaker, or equivalent).
Experience with experiment tracking (MLflow, W&B) and model registry practices.
Domain & Soft Skills**
Ability to translate ambiguous manufacturing/business problems into well-scoped AI solutions.
Strong analytical and structured problem-solving skills.
Excellent communication skills — able to present complex AI topics to non-technical stakeholders.
Intercultural collaboration experience; comfortable working in cross-functional, international teams.
Mandarin (working proficiency) and English (professional proficiency) required.
工作职责
Roles and Responsibilities
Agentic AI Development**
Identify agentic opportunities across RBCD manufacturing operations; translate business cases into technical agent architectures and implementation roadmaps.
Design, develop, and deploy AI agents using modern agentic frameworks (LangGraph, AutoGen, CrewAI, or equivalent) integrated into the M.AI.Co platform.
Leverage Coding Agents (Claude Code, OpenAI Codex, GitHub Copilot Workspace) to accelerate agent and tool development; apply prompt engineering, tool-use, and multi-agent orchestration patterns.
Architect multi-agent pipelines that combine LLMs, vision models, retrieval-augmented generation (RAG), and structured data to deliver end-to-end automated workflows.
Build and maintain safety guardrails: human-in-the-loop checkpoints, output validation, audit logging, and risk control frameworks for production agent deployments.
Manage agent data assets — define retrieval strategy, knowledge-base architecture, and context window optimization to ensure agents have accurate, up-to-date information.
Computer Vision & Classical AI**
Lead end-to-end development of computer vision and deep learning algorithms for AOI (Automatic Optical Inspection) and manufacturing quality processes — from PoC to serial implementation.
Design and implement model architectures (CNN, Transformer-based ViT, multimodal) for classification, detection, segmentation, and anomaly detection tasks.
Establish data collection, labelling, and data management strategies to enable fast model iteration and continuous performance improvement.
Collaborate with IT, machine, and product technical experts to define data pipelines, storage concepts, and MLOps practices.
Leadership & Knowledge Sharing**
Lead or accompany cross-functional project teams, taking full responsibility for costs, quality, and on-time delivery.
Coordinate AI work packages within RBCD manufacturing; cooperate with PS central departments to deploy and share standard solutions across plants.
Coach and mentor data enthusiasts and team members; drive AI competency build-up including Agentic AI literacy across the organisation.
Proactively monitor the latest AI developments (foundation models, agentic frameworks, coding agents) and evaluate their applicability to RBCD use cases.
AI 洞察
优缺点分析
优点
- 接触最前沿的Agentic AI技术，技能提升快
- 博世作为跨国巨头，平台稳定，资源丰富，接触真实工业场景
- 职位职责完整，从研发到部署，能积累全栈AI经验
- 需要同时掌握智能体、计算机视觉和MLOps，技能要求全面
- 需要持续学习最新AI技术，保持技术水平
- 适合有AI实战经验，热爱技术创新，并希望在工业场景应用大模型的工程师
缺点 / 挑战
- 工业场景复杂度高，跨部门协作挑战多
角色解读
- 沿技术专家路线深入，成为Agentic AI领域的架构师
- 向技术管理方向发展，领导更大规模的AI团队和项目
- 跨部门调动，在博世全球工厂推广标准AI解决方案
- 设计并开发基于LangGraph/AutoGen等框架的多智能体系统，集成LLM、视觉模型和RAG，实现制造流程自动化
- 负责计算机视觉和深度学习算法在AOI和质量管理中的应用，从PoC到量产
- 构建智能体的安全护栏，包括人工复核节点、输出验证和审计日志，确保生产环境稳定
- 领导跨职能团队，协调AI工作包，并在内部推广Agentic AI知识
- 熟悉LangGraph、AutoGen、CrewAI等智能体框架，以及Claude Code、GitHub Copilot等编码Agent
- 精通Python和PyTorch，掌握YOLO、ViT等视觉模型和OpenCV图像处理
- 具备MLOps实践经验，包括Docker、Kubernetes、MLflow和CI/CD
- 优秀的问题分解和沟通能力，能将复杂业务问题转化为AI方案
申请策略
- 了解博世RBCD的业务，尤其是制造质量场景
- 准备好展示如何将AI技术落地到工业现场
- 突出Agentic AI或LLM应用的项目经历，如使用LangGraph/AutoGen构建多智能体系统
- 展示计算机视觉项目的端到端交付，包括模型设计和部署
- 强调MLOps实践和容器化部署经验
- 体现跨团队协作和领导力，如带领项目小组
- 如果缺乏Agentic AI经验，可学习LangGraph和AutoGen，并构建小项目
- 熟悉一个编码Agent（如Claude Code或Copilot）提高开发效率
面试指南
- 使用STAR法则回答行为问题，突出角色、行动、结果
- 对技术问题，先明确问题背景，再阐述解决方案，最后说明效果
- 展示系统性思维：从业务需求出发，考虑技术选型、数据、部署、监控全流程
- 请描述一个你从零到一部署的Agentic AI系统，你是如何设计架构和保证安全性的？
- 在AOI视觉检测中，如何解决数据不均衡和缺陷样本不足的问题？
- 你会如何向非技术背景的利益相关者解释复杂AI系统的价值？
- 如何评估一个新的智能体框架是否适用于生产环境？
- 请举例说明你在团队中如何推动AI知识的普及
职位点评
77
综合评分
跨国巨头AI专家岗，技术前沿，发展空间大，但工作地点无锡且WLB未明确。
从薪资福利、成长空间、工作节奏和岗位方向综合评估，方便横向比较。
更适合这类人
适合追求技术前沿、渴望在工业场景落地AI并看重职业稳定发展的求职者。
表现最好
成长发展
相对薄弱
工作生活
薪资福利80
成长发展92
工作生活55
使命价值75
薪资福利
80较高
博世作为跨国巨头，提供稳定的平台和有竞争力的薪酬，但JD未披露具体福利。
薪资信号未披露（AI估算：25K-40K/月）
成长发展
92较高
岗位聚焦前沿Agentic AI和LLM技术，技能成长空间大，同时有内部知识分享和导师文化。
技术前沿前沿/新兴技术
技术栈LangGraph、AutoGen、CrewAI、Claude Code、OpenAI Codex、RAG、LLM、Computer Vision、PyTorch、MLOps、Docker、Kubernetes
成长机会coach、mentor、competency build up、knowledge sharing
业务类型ambiguous
工作生活
55较低
工作地点无锡，JD未明确远程或弹性工作，工作强度未知。
工作模式未明确
办公地点未明确
加班情况未提及（无法判断）
使命价值
75中等
博世致力于降低排放和加速新能源产品，岗位通过AI优化制造流程，具有一定社会价值。
行业发展稳定成熟行业
社会影响正向社会影响力较高
使命信号reduce emission、accelerate the market launch of new energy products
创新程度积极采用新技术
博世 的其他在招职位
相似职位推荐
Watch Jobs