- 地点
- 北京,中国
- 部门
- 数字技术/信息技术
- 编号
- R4037723
职责
- 设计、实现和优化磁共振成像(MRI)图像重建算法和软件,涵盖传统方法(快速傅里叶变换/并行成像/压缩感知)和基于深度学习的重建方法。
- 实现 MRI 应用程序功能,设计、实现和维护应用程序服务和 API。
- 使用 PyTorch/TF 构建和训练重建网络(Unet/VarNet/物理引导、基于评分)。
- 提升性能、可靠性、可观测性和开发者工具,增强 GPU/CUDA 加速、多线程和内存优化,以实现实时/近实时侦察。
- 处理 k 空间数据、校准、线圈组合、ESPIRiT 图谱、偏置校正、后处理。
- 编写简洁、可测试的代码;单元测试/集成测试;持续集成/持续交付;容器化和部署。
- 符合 DICOM/ISMRMRD 标准;与 Gadgetron/BART/内部平台集成。
- 与磁共振物理学家和临床医生合作开展实验、基准测试和产品化。
- 参与架构设计、技术路线图制定和跨团队项目协作。
- 参与敏捷流程:计划、估算、回顾和值班(根据需要)。
必备要求
- 计算机科学、电子工程、生物医学工程、医学物理、应用数学或相关领域的硕士或博士学位。
- 扎实的数学和信号处理基础(线性代数、优化、傅里叶/采样理论)。
- 精通现代 C++,并具备扎实的工程实践经验。
- 掌握磁共振成像重建基础知识:k 空间、采样轨迹(笛卡尔/螺旋/径向)、并行成像。
- 至少精通以下一项:优化/反问题(ISTA/FISTA/ADMM/PGD);并行成像(SENSE/GRAPPA/ESPIRiT);非笛卡尔坐标系重建(NUFFT、Toeplitz/阻塞 Toeplitz 方法);深度学习侦察(端到端或物理约束网络,即插即用)。
- 熟悉 Linux、Git、调试和性能分析。
- 熟悉前端技术,例如 JavaScript、HTML5、CSS、AngularJS、BackboneJS、EmberJS 和 KnockoutJS。
- 熟悉数据库工具(JPA、Hibernate、JDBC、Spring Data)。
- 熟悉规则引擎(Camel、Drools、JRules)和现代工具(Gradle、Maven、Git、SVN),能够熟练构建诸如 Web 服务和对现有表进行查询等功能。
- 了解精益/敏捷/极限编程等方法和实践。
- 具备敏捷环境下的工作经验,或接受过精益/敏捷方面的大门培训,了解并能够应用持续集成(CI)。
- 对技术选择对软件开发生命周期的影响有总体了解。
- 能够分解问题并估算开发任务所需时间。
- 了解技术格局,掌握当前技术趋势和新技术,为团队带来新想法。
- 展现出对项目客户价值主张的理解,并致力于为目标客户提供最佳价值主张。
- 学习组织愿景声明和决策框架,能够理解团队和个人目标/任务如何为组织愿景做出贡献。
- 能表达观点并提出清晰的理由。运用数据或事实证据来影响他人。
- 按时高质量地完成分配的任务。独立负责所分配的交付成果。
- 在实施解决方案之前,力求彻底了解问题。遇到歧义时,会提出问题以澄清需求。
- 发现创新机会并提出新想法。主动尝试新的软件框架。
- 能够适应新环境和不断变化的需求,并根据需要快速调整方向。在接受指导后,能够响应需求并从其他渠道获取信息,使用所选技术编写符合标准并实现所需功能的代码。
- 自我驱动、结果导向,将研究代码转化为生产代码。
- 强大的沟通能力和跨学科协作能力。
- 文档记录和实验可复现性思维。
- 美国以外地区的职位:计算机科学或“STEM”(科学、技术、工程和数学)专业学士学位,并具备基本经验。
优先条件
- CUDA/CuFFT/CUBLAS、OpenMP、SIMD 向量化。
- 熟悉 Git、Jenkins、DevOps 等技术。
- 亲身实践真实磁共振数据和序列(EPI、3D GRE/SE、VIBE、bSSFP、DWI/DTI、ASL、MRE)。
- DICOM/HL7、PACS 工作站集成和医疗软件合规性(IEC 62304、ISO 13485)。
- 出版物/竞赛(例如 fastMRI)或开源贡献。
Responsibilities
- Design, implement, and optimize MRI image reconstruction algorithms and software, covering traditional methods (FFT/parallel imaging/compressed sensing) and deep learning-based reconstruction.
- Implement MRI application features; design, implement, and maintain application services and APIs.
- Build and train reconstruction networks (Unet/VarNet/physics-guided, score-based) using PyTorch/TF.
- Improve performance, reliability, observability, and developer tools; enhance GPU/CUDA acceleration, multithreading, and memory optimization for real-time/near-real-time scouting.
- Process k-space data, calibration, coil combination, ESPIRiT maps, bias correction, and post-processing.
- Write clean, testable code; unit/integration tests; CI/CD; containerization and deployment.
- Comply with DICOM/ISMRMRD standards; integrate with Gadgetron/BART/in-house platforms.
- Collaborate with MR physicists and clinicians on experiments, benchmarking, and productization.
- Participate in architecture design, technology roadmap, and cross-team projects.
- Participate in agile processes: planning, estimation, retrospectives, and on-call (as needed).
Must Have
- Master's or PhD in Computer Science, Electrical Engineering, Biomedical Engineering, Medical Physics, Applied Mathematics, or related field.
- Solid foundation in mathematics and signal processing (linear algebra, optimization, Fourier/sampling theory).
- Proficiency in modern C++ with solid engineering practice.
- Knowledge of MRI reconstruction fundamentals: k-space, sampling trajectories (Cartesian/spiral/radial), parallel imaging.
- Proficiency in at least one of: optimization/inverse problems (ISTA/FISTA/ADMM/PGD); parallel imaging (SENSE/GRAPPA/ESPIRiT); non-Cartesian reconstruction (NUFFT, Toeplitz/block-Toeplitz methods); deep learning reconstruction (end-to-end or physics-constrained networks, plug-and-play).
- Familiarity with Linux, Git, debugging, and profiling.
- Familiarity with front-end technologies such as JavaScript, HTML5, CSS, AngularJS, BackboneJS, EmberJS, and KnockoutJS.
- Familiarity with database tools (JPA, Hibernate, JDBC, Spring Data).
- Familiarity with rule engines (Camel, Drools, JRules) and modern tools (Gradle, Maven, Git, SVN), able to build features such as web services and queries on existing tables.
- Knowledge of Lean/Agile/XP methods and practices.
- Experience in agile environments or training in Lean/Agile, with knowledge and ability to apply CI.
- General understanding of the impact of technology choices on the software development lifecycle.
- Ability to decompose problems and estimate development tasks.
- Understanding of the technology landscape, current trends, and new technologies to bring new ideas to the team.
- Demonstrated understanding of the project's customer value proposition and commitment to delivering the best value proposition.
- Learn organizational vision statements and decision frameworks; understand how team and individual goals/tasks contribute to the organizational vision.
- Ability to express opinions and provide clear rationale; use data or factual evidence to influence others.
- Complete assigned tasks on time with high quality; independently own assigned deliverables.
- Strive to thoroughly understand problems before implementing solutions; ask questions to clarify requirements when ambiguous.
- Identify innovation opportunities and propose new ideas; proactively try new software frameworks.
- Adapt to new environments and changing requirements; quickly pivot as needed. After receiving guidance, respond to requirements and obtain information from other sources, writing code that meets standards and achieves required functionality using chosen technologies.
- Self-driven, results-oriented, able to turn research code into production code.
- Strong communication and cross-disciplinary collaboration skills.
- Documentation and experimental reproducibility mindset.
- For positions outside the US: Bachelor's degree in Computer Science or STEM with basic experience.
Preferred
- CUDA/CuFFT/CUBLAS, OpenMP, SIMD vectorization.
- Familiarity with Git, Jenkins, DevOps, etc.
- Hands-on experience with real MR data and sequences (EPI, 3D GRE/SE, VIBE, bSSFP, DWI/DTI, ASL, MRE).
- DICOM/HL7, PACS workstation integration, and medical software compliance (IEC 62304, ISO 13485).
- Publications/competitions (e.g., fastMRI) or open-source contributions.
公司简介
GE 医疗是全球领先的医疗技术和数字化解决方案创新者,宗旨是创造一个医疗无界限的世界。我们重视包容性,是一家机会均等的雇主,所有雇佣决定均不考虑种族、肤色、宗教、国籍或民族血统、性别、性取向、性别认同或表达、年龄、残疾、受保护的退伍军人身份或其他受法律保护的特征。我们期望所有员工践行行为准则:谦逊行事,建立信任;以透明的方式领导;专注地完成工作,并积极承担责任——始终坚持坚定不移的诚信。我们提供全面的薪酬福利,旨在激发您的雄心壮志,为您提供所需的动力和灵活性,助您将想法转化为改变世界的现实。