与我们一起讨论行业的今天与未来

推荐文章

才云+Kubeflow

才云科技2017-12-21

      才云是中国领先的容器云和深度学习云服务提供商。团队核心成员来自美国谷歌、亚马逊、卡耐基梅隆大学顶尖云计算和机器学习团队。才云主要产品包括容器集群智能云平台 Compass 和分布式机器学习平台 Clever。通过容器集群技术为企业提供 IT 开发、测试、部署、生产、智能监控、运维等全链条、全自动的云端服务;通过大规模分布式机器学习平台,结合海量数据与人工智能解决方案,助力企业科技转型、激活产品不断创新、为用户带来宝贵商业价值。

Caicloud is a leading player in Chinese Kubernetes and Deep Learning Cloud Ecosystem. With its initial members from Google, Amazon AWS, Carnegie Mellon University, etc. background, the core team empowers two major product lines: Compass, the enterprise containerization solution, and Clever, the one-stop platform for distributed machine learning applications. Compass is our answer for enterprise scale IT demands from development, testing, deployment, service, monitoring, to operations, with a focus on full automation and full autonomy; Clever, on the other hand, is enabling business values for our clients through distributed machine learning platform, connecting big data with AI solutions, propelling the transformation of enterprises towards a more digitized and intelligent vision.

      才云在电商、金融、运营商、教育、传统企业等行业均有成熟解决方案,并且已在众多大型 500 强企业中落地运行。如:TOP 3 大型银行的智慧网点中的人脸识别,人卡证合一,动线分析,陌生获客;著名网红电商新零售中的衣物、风格、面料分类,以及智能推荐等。

With our clients spreading across e-commerce, financial, telecommunications, educational tech, as well as more traditional manufacturing industries, Caicloud has already established successfully operational cases amongst many Fortune 500 companies. ABC, one of the top three banks in China, has built their Smart Bank with our face recognition solution, and can now analyze customers behavior and better engage new customers; A famous internet-celebrity e-commerce company has deployed our image search engine to help their fashion designers to sort, organize, and discover photos of different styles, textures, etc. for better inspirations.

      才云一直秉承「以客户为中心」的价值理念,不断打磨产品来满足客户对人工智能解决方案日益增长的实际需求。同时,才云也积极参与开源社区,我们拥有数位顶尖的 Kubernetes 社区 Members 和 TensorFlow 社区的活跃贡献者。

Putting client first has always been our core values, and with that we continuously iterate our products in order to meet our clients’ ever growing needs for AI solutions. In the meantime, we have also devoted to giving back to open source communities. Many of the core members of Kubernetes community and active contributors in TensorFlow community are from Caicloud team.

      从才云的用户反馈和全球 AI 发展的趋势来看,Machine Learning on Kubernetes 是一个实际需求。放眼未来,Machine Learning on Kubernetes 将成为 AI 基础设施。一个典型的端到端 AI 解决方案应该包括:商业需求-产品设计-数据收集-数据处理-模型训练-模型服务-产品验证-商业变现。为了帮助 AI 工程师摆脱繁琐的环境搭建和资源配置工作,更聚焦于数据和模型,才云正和谷歌,以及社区贡献者们一同打造全新的开源项目——kubeflow。

There’s little doubt that Machine Learning on Kubernetes is a trend with a clear future, and that future includes our clients’ needs and the zeitgeist of global AI industry. In no time, this trend towards Kubernetes will establish as a cornerstone for AI, and especially Machine Learning. Upon that foundation, a typical AI solution should include the cycle of business need, product design, data collection, data processing, model training, model serving, production verification, and then business success. Our goal here is to free AI engineers from the cumbersome setup of environment and resource management, and focus on what they are good at: data and model. Kubeflow, the new open source project with its name coined from combination of two communities, is initiated by Google for this very purpose, and Caicloud, along with other early contributors, will continue to invest in this endeavor.

      Kubeflow 旨在利用 Kubernetes 强大的资源管理和调度能力,为所有主流的 Machine Learning 框架提供平台级支持。Kubeflow 的设计目标清晰,但想要解决的问题并不简单。这其中涉及到很多先进技术的容器化,典型的包括:

  • 对多种硬件加速资源的细粒度识别、管理和调度;

  • 对多种机器学习框架的深度支持;

  • 对大数据计算负载的深度支持;

  • 对高性能计算负载的深度支持;

Kubeflow is targeted at leveraging the strong and already proven scheduling and management ability of Kubernetes, to support mainstream machine learning frameworks as a platform. The goal is clear, but the challenges remain non-trivial, because it involves containerization of multiple modern technologies, e.g.

  • fine-grained identification, management, and scheduling of multiple hardware-accelerated resources

  • integrated support for different kinds of machine learning frameworks

  • integrated support for calculation on massive datasets

  • integrated support for high performance computing

     事实上,才云已经实现了许多相关工作,我们也将会分阶段地把该技术回馈给开源社区。

In fact, Caicloud has already heavily invested in some aspects of them, and we are looking forward to open sourcing these technologies in near future.

      早在2017年初,才云便已打造出基于 Kubernetes 和 TensorFlow 的成熟产品——分布式深度学习平台 TaaS(TensorFlow as a Service)。一经面世,TaaS 便受到国内外广泛关注。在产业界,我们将 TaaS 和时尚、安防、教育、零售相结合,催生出一系列成功的 AI+ 产品。在学术界,我们于 2017 年 3 月柏林 KubeCon 上隆重推出了 TaaS 平台的设计理念与实践经验。

As early as 2017 Spring, Caicloud has revealed a full-grown product based on Kubernetes and TensorFlow: TensorFlow-as-a-Service, a.k.a. TaaS. Upon publication, TaaS has attracted wide attention from domestic and international markets. In industrial cases, we applied TaaS to fashion, security, education, and FMCG businesses, where we showcased multiple successful AI+ products, and shared our design philosophies and best practices with the community in the Berlin KubeCon in March 2017.

      才云非常愿意与开源社区一起共同推进容器技术和人工智能技术的进步,把两者有机结合将进一步提升人类文明的智能化水平。希望才云能够作为容器开源社区和人工智能开源社区的桥梁和中流砥柱,为 kubeflow 和全人类做出卓越的贡献。

To conclude, Caicloud team is very happy to dash forward with the community towards the improvement of container and AI technologies, and the synergy between these two for a more clever and intelligent world. There’s more to come in Kubeflow and other open source projects, and we are happy to play an important part of it.