1. Large Model Data Production Platform
- Build a data production platform for data aggregation, automated cleaning, and corpus synthesis, targeting workflows for researc……
1、大模型数据生产平台 -面向研究员及研发工程师的工作流，建设数据汇聚、自动清洗、语料合成等数据生产平台。 -针对不同数据源及任务流，设计并研发自动化数据处理流程，支撑海量数据的高效生产与处理。 -协同算力基础设施，持续优化数据处理效率、稳定性及资源利用率。 2、AI 研发效能与 IT 能效 -结合人员、项目及算力资源，建设研发效能与运行管理平台，提升 AI 研发过程的协同效率。 -打通项目管理、研发过程、算力资源及任务执行等关键环节，形成端到端的 AI 生产管理能力。 -持续建设研发度量、资源分析及运行监控能力，为研发效率和资源利用率优化提供数据支撑。 3、算力潮汐调度与运行管理 -对接 MaaS 平台及算力调度平台，面向离线训练、数据处理等任务建设算力调度能力，充分利用集群闲时资源。 -面向大规模算力集群，参与建设统一作业管理、训练编排、实验管理、数据管理等核心能力。 -持续优化任务调度、资源分配及集群运行效率，提升整体算力利用率和资源产出。 1、硕士及以上学历，计算机等相关专业，3 年及以上研发经验，有云原生、分布式计算或 AI 基础设施相关经验。 2、具备扎实的工程能力，熟悉 TypeScript、Shell，具备良好的系统设计与代码实现能力。 3、熟悉 Git 等代码管理工具，具备良好的工程规范和代码质量意识，熟练掌握 Linux 开发环境。 4、深入理解 Kubernetes 架构及核心机制，具备 Kubernetes 二次开发或平台研发经验。 5、熟悉 Volcano、Kueue、Ray 等调度或分布式计算框架之一；具备大规模分布式系统研发经验者优先。 6、具备良好的 Vibe Coding 能力，能够熟练运用 AI 编程工具提升研发效率，同时对技术方案、工程质量和代码品质有较高要求。 具备良好的沟通协作能力、责任心和自驱力，能够主动推动跨团队技术问题解决，具备较强的学习和技术演进能力。
🈶
中文翻译 · AI 生成，以原文为准
1. Large Model Data Production Platform - Build a data production platform for data aggregation, automated cleaning, and corpus synthesis, targeting workflows for researchers and R&D engineers. - Design and develop automated data processing pipelines for different data sources and task flows to support efficient production and processing of massive data. - Collaborate with computing infrastructure to continuously optimize data processing efficiency, stability, and resource utilization. 2. AI R&D Efficiency and IT Energy Efficiency - Build an R&D efficiency and operations management platform that integrates personnel, projects, and computing resources to improve collaboration efficiency in the AI R&D process. - Connect key stages such as project management, R&D processes, computing resources, and task execution to form end-to-end AI production management capabilities. - Continuously build R&D measurement, resource analysis, and operations monitoring capabilities to provide data support for optimizing R&D efficiency and resource utilization. 3. Computing Power Tidal Scheduling and Operations Management - Integrate with the MaaS platform and computing power scheduling platform to build computing power scheduling capabilities for tasks such as offline training and data processing, fully utilizing idle cluster resources. - For large-scale computing clusters, participate in building core capabilities such as unified job management, training orchestration, experiment management, and data management. - Continuously optimize task scheduling, resource allocation, and cluster operation efficiency to improve overall computing power utilization and resource output. 1. Master's degree or above in Computer Science or related fields, 3+ years of R&D experience, with experience in cloud-native, distributed computing, or AI infrastructure. 2. Solid engineering skills, familiar with TypeScript and Shell, with strong system design and code implementation abilities. 3. Familiar with code management tools such as Git, with strong awareness of engineering standards and code quality, and proficient in the Linux development environment. 4. Deep understanding of Kubernetes architecture and core mechanisms, with experience in Kubernetes secondary development or platform R&D. 5. Familiar with one of the scheduling or distributed computing frameworks such as Volcano, Kueue, or Ray; experience in large-scale distributed system R&D is preferred. 6. Strong Vibe Coding ability, able to proficiently use AI programming tools to improve R&D efficiency, while maintaining high standards for technical solutions, engineering quality, and code quality. Strong communication and collaboration skills, sense of responsibility, and self-motivation; able to proactively drive cross-team technical problem resolution, with strong learning and technical evolution capabilities.