DoorDash正在从microrepos转向monorepo,以简化依赖管理和代码重用。尽管monorepo带来了可扩展性和构建时间的挑战,但DoorDash通过自研的分布式构建服务解决了这些问题,实现了高效的并行构建。该服务使用Gradle,支持Kotlin和protobuf项目,通过动态配置和资源管理优化构建过程,显著提高了构建速度和可靠性,提升了开发者的工作效率。

DoorDash采用大语言模型优化搜索系统,通过知识图谱和混合检索策略提升查询理解。系统灵活处理复杂查询,精准链接用户意图与商品属性,确保搜索结果相关性。结合检索增强生成技术,有效减少模型幻觉,提高检索精度。在线测试显示,新系统显著提升用户体验和转化率,为动态市场环境下的复杂查询处理提供了新思路。

DoorDash通过机器学习与生成式AI融合,提升客户体验,扩展至杂货、酒精等新领域。ML模型优化产品推荐、配送路线等,LLM助力产品知识图谱构建,提升搜索精准度。LLM辅助标注加速数据处理,自动化检测产品目录不一致,确保购物体验流畅。多意图、多实体、地理位置感知的搜索引擎,结合个性化推荐,提供更相关、易用的搜索结果。DoorDash持续探索多模态LLM,提升服务个性化与效率。

DoorDash通过使用大型语言模型(LLMs)和检索增强生成系统(RAG)优化了其Dasher支持系统。新系统能快速解决配送问题,但面临响应准确性、语言一致性等挑战。引入LLM Guardrail和LLM Judge等工具确保响应质量和合规性。尽管自动化提高了效率,复杂问题仍需人工处理。未来,系统将持续改进,以增强处理复杂问题的能力。

DoorDash leverages interleaving designs to boost experimentation sensitivity, enabling faster, more precise insights compared to traditional A/B testing.

Learn how DoorDash built a Kafka self-serve platform to streamline infrastructure management, empower engineers, and enhance scalability and security.

Discover how DoorDash uses metric-aware rollouts to enhance app health and user experience while innovating product development strategies.

Discover how DoorDash implemented a multi-tenancy awareness system for both Kafka producers and consumers.

DoorDash proactively embeds privacy into its products by geomasking user address data while maintaining local analytic capabilities.

In this post, we will describe how we built an eBPF-powered agent to monitor networking traffic within our Kubernetes clusters.

DoorDash's evolution from restaurant deliveries to a diverse business portfolio is powered by a robust logistics platform.

DoorDash leverages Large Language Models to build a product knowledge graph that captures relationships between millions of consumer goods.