公司：Uber
优步（英语：Uber，/ˈuːbər/）是一间交通网络公司，总部位于美国加利福尼亚州旧金山，以开发移动应用程序连结乘客和司机，提供载客车辆租赁及媒合共乘的分享型经济服务。乘客可以透过应用程序来预约这些载客的车辆，并且追踪车辆的位置。营运据点分布在全球785个大都市。人们可以透过网站或是手机应用程序进入平台。
优步的名称大多认为是源自于德文über，和over是同源，意思是“在…上面”。 （页面存档备份，存于互联网档案馆）
然而其营业模式在部分地区面临法律问题，其非典型的经营模式在部分地区可能会有非法营运车辆的问题，有部分国家或地区已立法将之合法化，例如美国加州及中国北京及上海。原因在于优步是将出租车行业转型成社群平台，叫车的客户透过手机APP（应用程序），就能与欲兼职司机的优步用户和与有闲置车辆的租户间三者联系，一旦交易成功即按比例抽佣金、分成给予反馈等去监管化的金融手法。
2019年5月10日，优步公司透过公开分发股票成为上市公司，但首日即跌破分发价。
据估算，优步在全球有1.1亿活跃用户，在美国有69%的市占率。优步亦在大中华区开展业务，目前优步已在香港和台湾建成主流召车服务平台，并于中国大陆通过换股方式持有该市场最大网约车出行平台滴滴出行母公司小桔科技17.7%经济权益。
Uber’s Journey to Ray on Kubernetes: Ray Setup
Uber将机器学习工作负载迁移至Kubernetes，解决了资源管理复杂、利用率低和容量规划不灵活等问题。通过构建联邦资源管理架构，抽象化底层基础设施，实现自动化资源分配和优化。新架构显著提升了训练速度，并提高了GPU资源的利用率，最终在2024年初完成所有项目的迁移。
Enhancing Personalized CRM Communication with Contextual Bandit Strategies
在CRM领域，优化邮件和推送通知对提升用户参与度至关重要。Uber采用上下文多臂老虎机方法，动态调整邮件内容，结合用户偏好和上下文特征，通过GPT嵌入和XGBoost模型预测最佳邮件变体，提升打开率。相比传统A/B测试，该方法能快速收敛，适应动态环境，显著提高个性化沟通效果，增强用户互动。
Automating Efficiency of Go programs with Profile-Guided Optimizations
Profile-guided optimization (PGO) 通过运行时数据指导编译器优化，提升代码性能。Uber 将 PGO 集成到持续优化框架中，通过日常数据收集、服务注册、CI 测试和部署，显著减少了资源消耗。针对编译时间增加的挑战，Uber 开发了预处理工具，大幅缩短了编译时间。PGO 在 Go 语言中的支持从 v1.20 开始，通过优化如函数内联和代码布局，实现了 4% 的性能提升，减少了 24,000 个 CPU 核心的使用。
Adopting Arm at Scale: Transitioning to a Multi-Architecture Environment
Uber在x86架构基础上成功引入Arm架构，通过调整代码库、更新容器镜像和优化构建工具链，逐步将5000多项服务迁移至Arm主机。采用分层、分区的策略，先在非生产环境测试，再逐步推广到生产环境，确保服务稳定性和性能。最终，2800多项无状态Go服务成功迁移，20%的Oracle云基础设施从x86转向Arm，提升了性价比并降低了碳足迹。未来将进一步扩展至Java服务和有状态工作负载。
Adopting Arm at Scale: Bootstrapping Infrastructure
Uber在2023年启动了一项战略迁移，将本地数据中心转移到Oracle和Google云平台，同时引入Arm架构主机以降低成本、提升性能并增强硬件灵活性。迁移过程中，Uber克服了从主机兼容性到构建多架构容器镜像的技术挑战，逐步将服务迁移至Arm环境。这一举措不仅优化了能源效率，还为未来的硬件多样性奠定了基础。
MySQL At Uber
Uber的MySQL数据库系统是其数据基础设施的核心,管理着2300多个独立集群,确保高可用性和数据无丢失。通过控制平面的重新设计,实现了从99.9%到99.99%的可用性提升。关键功能包括集群管理、主节点故障切换、节点生命周期管理和数据库操作。自动化流程如节点替换和架构变更,确保了系统的灵活性和稳定性。使用Cadence驱动的工作流提升了操作效率,提供了可靠的MySQL平台支持Uber的多种服务需求。
How Uber Uses Ray® to Optimize the Rides Business
Computational efficiency is a significant challenge when scaling solutions to a marketplace as large and as complex as Uber. The running and tuning of the Uber rides business relies on substantial…
Serving Millions of Apache Pinot™ Queries with Neutrino
Apache Pinot™ is a real-time OLAP database capable of ingesting data from real-time streams and offline data sources. Uber has used Pinot for over 6 years, and in that time we’ve seen a rapid increase…
Introducing the Prompt Engineering Toolkit
A well-crafted prompt is essential for obtaining accurate and relevant outputs from LLMs (Large Language Models). Prompt design enables users new to machine learning to control model output with…
The Accounter: Scaling Operational Throughput on Uber’s Stateful Platform
In a previous post, we introduced Uber’s stateful platform, Odin. We discussed how the platform’s scale and the growing need for fleet-wide operations required better coordination among its many…
Unified Checkout: Streamlining Uber’s Payment Ecosystem
Uber started as a ridesharing company with a simple yet ambitious mission: make transportation as reliable as running water, everywhere, for everyone. When the company focused on ridesharing, most…
Presto® Express: Speeding up Query Processing with Minimal Resources
Presto® is an open-source, distributed SQL query engine designed for running interactive analytic queries on data sources of any size, from gigabytes to petabytes. At Uber, Presto is a critical engine…
Enabling Infinite Retention for Upsert Tables in Apache Pinot
Apache Pinot™ was originally designed as an append-only OLAP (online analytical processing) database. After some redesign, it was modified to support upserts, which are UPdates plus inSERTs. This…
Streamlining Financial Precision: Uber’s Advanced Settlement Accounting System
In the realm of fintech, settlements, refunds, and chargebacks related to PSPs (payment service providers) are crucial transactions that involve the transfer of funds between different financial…
Open Source and In-House: How Uber Optimizes LLM Training
Generative AI powered by LLMs (Large Language Models) has a wide range of applications at Uber, like Uber Eats recommendations and search, customer support chatbots, code development, and SQL query…
Genie: Uber’s Gen AI On-Call Copilot
In today’s fast-paced tech environment, maintaining robust on-call operations is crucial for ensuring seamless service functioning. Modern platform engineering teams face the challenge of efficiently managing on-call schedules, incident response, communication during critical moments, and strong customer support on Slack® channels.
This post describes Genie, an on-call copilot we built that uses generative AI to optimize communication and question-answering with on-call engineers.