岗位描述
1) Computer Vision Algorithm R&D | 计算机视觉算法研发 Research, develop, and validate computer vision models for fine-grained recognition in complex real-world conditions. 在复杂真实环境下，研发并验证用于细粒度识别的计算机视觉算法。 Build training and evaluation of pipelines, including data preparation, augmentation, robustness testing, and error analysis. 构建训练与评估流程，包括数据准备、增强策略、鲁棒性测试与误差分析。 2) Edge AI Optimization & Deployment | 端侧 AI 优化与部署 Design lightweight, high-efficiency deep learning models for embedded/edge platforms. 面向嵌入式/端侧平台设计轻量高效的深度学习模型。 Optimize and deploy models using quantization / pruning / distillation, meeting constraints on latency, memory, and power. 通过量化/剪枝/蒸馏等方法优化并部署模型，满足时延、内存、功耗等约束。 Conduct on-device profiling and performance tuning with common inference toolchains. 基于主流推理工具链完成端侧profiling 与性能调优。 3) Multi-sensor Fusion | 多传感器融合 Fuse vision with other sensor modalities (e.g., inertial/vibration, load, spectral) to improve performance and reliability. 融合视觉与其他传感模态（如惯导/振动、载荷、光谱等），提升性能与可靠性。 Design fusion strategies and run ablation studies to quantify sensor contributions and failure modes. 设计融合策略并开展消融分析，量化传感器贡献与失效模式。 4) System Co-design, Innovation & Transfer | 系统共设、创新与转化 Collaborate with hardware and product teams to define camera/optics specifications and overall system architecture. 与硬件及产品团队协作，定义摄像头/光学规格与系统架构。 Scout emerging technologies, generate patents/publications, and transfer research into production-ready solutions. 跟踪前沿技术，产出专利/论文，并推动成果走向可量产方案。 Job Qualification: Computer Vision & Deep Learning | 计算机视觉与深度学习 Hands-on experience in at least one of: detection / segmentation / fine-grained classification / texture & material recognition. 至少在以下一项有落地经验：检测 / 分割 / 细粒度分类 / 纹理与材质识别。 Proficient in PyTorch or TensorFlow; capable of training, tuning, evaluation, and reproduction. 精通 PyTorch 或 TensorFlow；具备训练、调参、评估与复现能力。 Solid data & experimentation practice: data cleaning, annotation strategy, augmentation, bias/robustness analysis. 具备数据与实验规范：数据清洗、标注策略、数据增强、偏差与鲁棒性分析。 Embedded/Edge AI Deployment & Optimization | 端侧部署与模型优化 Proven deployment experience on at least one platform: ARM / DSP / NPU. 具备至少一种平台的部署经验：ARM / DSP / NPU。 Familiar with model optimization: quantization (PTQ/QAT), pruning, distillation, mixed precision, structural re-parameterization. 熟悉模型优化：量化（PTQ/QAT）、剪枝、蒸馏、混合精度、结构重参数化。 Experience with at least one inference/compile toolchain: TFLite/TFLite Micro, TensorRT, TVM, ONNX Runtime, vendor NPU SDK. 熟悉至少一种推理/编译工具链：TFLite/TFLite Micro、TensorRT、TVM、ONNX Runtime、厂商 NPU SDK。 Performance engineering: profiling latency/throughput/memory/power, bottleneck analysis (operator-level preferred). 具备性能工程能力：profiling 延迟/吞吐/内存/功耗，并能定位瓶颈（算子级分析优先）。 Multi-sensor Fusion & Signal Processing | 多传感器融合与信号处理 Experience in sensor fusion: Kalman/Particle filtering or multi-modal neural fusion. 有融合经验：卡尔曼/粒子滤波或多模态神经网络融合。 Signal processing capability for time-series/sensor signals: FFT/STFT, time-frequency analysis, statistical features. 具备时序/传感信号处理能力：FFT/STFT、时频分析、统计特征提取。 Engineering & Implementation | 工程实现能力 Strong coding in Python & C/C++, with maintainable engineering practices (modularization, testing, optimization). 扎实的 Python 与 C/C++ 编程能力，具备可维护工程习惯（模块化、测试、优化）。 Familiar with embedded/edge dev & debugging workflow: cross-compiling, logging/tracing, HIL testing. 熟悉端侧开发与调试流程：交叉编译、日志/追踪、硬件在环测试。 Soft Skills | 通用能力 Strong analytical and problem-solving ability; able to drive end-to-end delivery. 出色的分析与问题拆解能力，能推动端到端交付。 Good communication in both English & Chinese (written and verbal) for global collaboration. 良好的中英文沟通能力（书面与口头），适应全球协作。 Preferred Qualifications (Bonus) | 加分项 Deep expertise in texture analysis / material recognition / fine-grained recognition (industrial scenes preferred). 在纹理分析 / 材质识别 / 细粒度识别方面有深入积累（工业场景优先）。 Multi-modal perception experience: fusion of vision + acoustic/vibration/spectral/IMU signals. 多模态感知经验：融合视觉 + 声学/振动/光谱/惯导等信号。 Optical & illumination design: optical path/lighting design, lens selection, imaging system setup & calibration. 光学与照明设计：光路/照明设计、镜头选型、成像系统搭建与标定。 Spectral analysis: spectral sensors, preprocessing & calibration, feature extraction, quantitative/semi-quantitative modeling. 光谱分析：光谱传感器应用、预处理与校准、特征提取、定量/半定量建模。 Camera imaging pipeline knowledge: ISP tuning / image quality optimization (AE/AWB/NR/HDR, etc.). 摄像头成像链路：ISP 调试 / 画质优化（AE/AWB/降噪/HDR 等）。 Publications or high-quality patents in top venues (CVPR/ICCV/ECCV/NeurIPS, etc.). 顶会/顶刊论文或高质量专利产出（CVPR/ICCV/ECCV/NeurIPS 等）。 Experience in home appliances / smart home product development (e.g., washing machines, refrigerators, dishwashers), from algorithm R&D to productization. 具备家电/智能家居产品研发经验（如洗衣机、冰箱、洗碗机等），有从算法研发到工程落地/产品化的经验。 博世集团网上招聘数据处理同意声明 个人信息及目的：你的如下个人信息将被处理并用于招聘：姓名、电子邮箱、电话号码、所在地区、经验和教育信息、简历等应聘者自主提交的附件信息、证件号、职位、部门、计划入职日期、劳动合同持续时间、工作地点、月薪、年薪、浮动奖金比例、餐贴、车贴。 博世将依据相关的数据保护法规，在全球范围内对您的数据进行保密。招聘流程中，您的个人信息只有在必须时才会被提供给博世以外的第三方机构(如代理机构)。博世精心挑选了第三方机构并与之签订了保密合同。博世采取了种种保密措施以保护您的数据，确保其不会遭受操纵、丢失、破坏、未经授权的访问或泄漏。我们的安全措施将随着新技术的应用而不断升级。您的本地浏览器和我们的人才招募系统之间的数据传输通过https加密。 特定职位申请：申请博世集团的职位