Luis Miguel Benitez Ruiz

我参与构建了一个数据分析管道，该管道编排了多个服务，以自动注释亚马逊服装目录中的商品. 然后，我使用MxNet网络自动预测，并从目录图像中提取有意义的属性.g., floral pattern, scoop neck), which were used to enhance the catalog items, 这样亚马逊的顾客在购买服装时就可以用它们来搜索或过滤.

I was involved in building the prediction services that used MxNet, 快速注释和审查服务:一个web应用程序，数据注释者可以审查低置信度的预测. Also built a data warehouse to collect metrics throughout the whole process. 我们的数据科学家用它来改进现有的模型并进行新的实验.

作为DevOps团队的一员，我开发了一个随叫随到的轮换监控工具. 这个应用程序监控我们的bug跟踪数据库，并允许工程师为他们选择的时间框架和产品创建随叫随到的班次, set up alerts to their email and mobile number, and allow managers to set up escalation rules for high-severity incidents.

Before this, the organization had no tool for managing on-call, so rotation was controlled differently across teams, mostly through shared documents or emails. 设置警报很复杂，并且需要工程师在他们通常不应该访问的系统中拥有特权.

一个集中的随叫随到的工具为所有团队的随到轮换提供了更好的可见性和问责制, made the onboarding process easy and quick for new team members, and helped keep sensitive systems more secure.

它是如此的受欢迎，以至于一开始只是我团队的一个个人项目，很快就变成了一个新的内部产品, part of the standard tools for the entire organization.

合作开发一个web应用程序，允许通过使用在线机器学习对非常大的数据集进行快速手动标记.

当用户提供标记数据时，应用程序迭代地构建分类模型并生成质量度量. The annotator had a very streamlined and simple UX, 数据科学家可以访问生成模型的聚合指标和其他元数据.

With this tool, annotators had to label only from 1% to 4% of the samples in a dataset, while before, they had to manually label 100% of the dataset. 这使得在为我们的数据科学团队生成高质量的标记数据集时可以更快地周转.

CloudFix is a cost-optimization service for AWS. It scans AWS resources in an organization, identifies inefficient resources, and automatically and securely fixes them.

I collaborated in designing the architecture of the CloudFix framework, designing the detection and fixing algorithms, and implementing the product in TypeScript and Node.js.