自动标注算法工程师（模型方向）_XC at Bosch Group

Shanghai, Shanghai, China -

Full Time

Start Date

Immediate

Expiry Date

17 Feb, 26

Salary

0.0

Posted On

19 Nov, 25

Experience

2 year(s) or above

Remote Job

Yes

Telecommute

Yes

Sponsor Visa

Skills

Static Pre-Labeling, Dynamic Pre-Labeling, Model Training, Model Optimization, Data Processing, LiDAR, Camera, Multi-Modal Data, Cloud Inference Pipeline, Tracking, Quality Control, Data Mining, Semi-Supervised Learning, Weak Supervision, Pseudo-Labeling, Model Deployment

Industry

Software Development

Description

公司介绍 Do you want beneficial technologies being shaped by your ideas? Whether in the areas of mobility solutions, consumer goods, industrial technology or energy and building technology - with us, you will have the chance to improve quality of life all across the globe. Welcome to Bosch. 职位描述 A. 静态预刷（Static Pre-Labeling）模型负责云端静态自动标注模型的设计、训练与迭代，包括但不限于：车道线、路沿、路面箭头、交通标志等静态要素的特征建模与预测；针对道路拓扑（Topology）构建 merge / split 点识别、入口/出口关系建模、拓扑一致性校验；利用多模态数据（LiDAR intensity, RGB Map, Semantic Map 等）提升静态精度；设计端到端或模块化静态感知模型例如 VMA 等；构建静态预刷模型的版本管理、评测体系与大规模自动化训练流程。 B. 动态预刷（Dynamic Pre-Labeling）模型根据业务需求承担动态预刷模型的训练与开发，包括：负责目标检测、跟踪（Tracking）、轨迹预测等动态感知模型训练与优化；具备处理多车型数据（轿车/卡车/专用车等）和跨车型泛化能力；掌握多传感器对齐与融合（Camera 和 LiDAR 之间，Front LiDAR, Main Lidar, Blind LiDAR 之间，Pinhole Camera, Fisheye Camera），解决异源数据问题；设计可迁移至大规模数据生产的 cloud inference pipeline；与 Tracking、QA、数据挖掘团队协作，实现动态标注数据的连贯性、准确性与模型驱动的自我演化。 C. 系统与闭环支持构建自动标注数据闭环（model → pre-label → QC → feedback → retrain）；分析模型误差来源，制定半监督 / 弱监督 / 伪标签（pseudo-label）策略；优化推理速度，提升大规模云端标注吞吐效率。职位要求硕士及以上学历，计算机视觉 / 自动驾驶 / AI 相关专业；扎实掌握 LiDAR/Camera 检测/分割等核心感知算法；有实际模型训练与调优经验，包括 loss 设计、数据增强、模型评测；理解自动驾驶数据结构、Frame/Clip、Pose、点云特性、Label 规范；具有工程落地能力，能处理大规模训练数据、分布式训练、模型部署。加分项（Highly Preferred）有静态感知经验：Lane/Boundary/Topo 建模、vector map learning；有动态感知经验：CenterPoint、BEVFusion、DeepSORT/ByteTrack、trajectory modeling；熟悉多车型适配 / 跨车型泛化相关建模经验；熟悉 Camera + LiDAR + Radar 融合，具备异源数据对齐经验；有自动标注产线、云端预刷系统或数据生产平台经验；熟悉半监督、弱监督、伪标签生成、多模型融合；有顶会论文或开源项目贡献（CVPR/ICCV/NeurIPS等）优先。

Responsibilities

The role involves designing, training, and iterating on static and dynamic pre-labeling models for automated labeling in cloud environments. Responsibilities include building models for various static elements and optimizing dynamic perception models for tracking and trajectory prediction.