Airbnb developed a production traffic capture and replay system to test database performance using real workloads instead of synthetic tests.
Platform for building reproducible e-commerce testing environments where shopping agents can be consistently evaluated against realistic online stores.
Deliver customized user experiences using behavioral signals and heuristics instead of explicit user identification.
Secure workloads on managed compute by replacing cloud provider identity with cryptographic attestation proving the actual running code.
An internal tool leveraging LLMs to assist with cross-platform mobile app migration while maintaining code quality through checkpoints and validation gates.
Post discusses moving beyond classpath-based dependency management toward modern approaches that improve system flexibility and deployment efficiency.
Airbnb enhanced their recommendation engine by integrating Chronon to serve freshly computed features that update in real time, improving personalization accuracy.
Building production AI infrastructure requires systematic thinking about data quality, monitoring, and operational concerns beyond just model performance.
Team rebuilt a mobile app using native technologies instead of cross-platform framework, achieving production launch in three months with AI assistance.
Shopify shifted from React Native to native Swift/Kotlin development due to changed economic incentives from AI coding agents affecting mobile app development costs.
Shopify's AI agent automates vulnerability remediation by autonomously fixing security issues and submitting them for review rather than just identifying problems.
Netflix developed a multimodal personalization system for efficiently delivering customized content assets at massive scale across diverse user devices.
A tool for anonymizing sensitive data in production systems, enabling safe analysis and debugging without exposing personally identifiable information.
Comparison of two autoscaling approaches for Apache Flink, examining their tradeoffs in handling dynamic workload changes and resource optimization.
How prediction models failed to adapt when pandemic conditions shifted, requiring teams to rebuild assumptions and detection mechanisms.
Technique for compressing LLM agent context into learned tokens to reduce latency and operational costs while maintaining quality.
Rebuilt mobile end-to-end testing infrastructure using structured APIs and computer vision to achieve 98% stability in automated test execution.
Airbnb redesigned their authentication infrastructure to handle millions of users with flexible, scalable mechanisms supporting diverse identity verification methods.
Netflix built a distributed graph system with gRPC-based querying to efficiently retrieve real-time data across their infrastructure.
Building a continuous learning system that incorporates production failures into model updates daily, achieving superior performance while dramatically reducing infrastructure costs.
Design a system to track and manage diverse device capabilities for analytics purposes across heterogeneous infrastructure.
Netflix developed a recommendation system leveraging large language models to generate personalized content suggestions natively through language modeling techniques.
Designing an AI-driven code review system that detects security vulnerabilities, validates them through testing, and generates fixes abstracted from underlying model changes.
Strategies for systematically evaluating generative AI systems in production, focusing on measurement frameworks and scaling evaluation processes.
Airbnb improved search relevance by analyzing guest behavior patterns and using machine learning to personalize listing recommendations throughout the booking journey.
Netflix built infrastructure to deploy and serve large language models efficiently within their own data centers rather than relying on external APIs.
Migrating checkout extensions to remote-dom and web components dramatically reduced bundle sizes and improved checkout performance at scale.
Optimizing LLM evaluation pipelines to reduce iteration time from weeks to a single day through performance improvements and infrastructure scaling.
A Shopify team rapidly prototyped a music player with custom visualizers and artist tools for store pages during a three-day hackathon.
Designing and managing service interdependency mappings across large microservice ecosystems, addressing discovery, consistency, and operational challenges.
Netflix developed an end-to-end generative system for personalized homepage construction, optimizing content ranking and layout generation at massive scale.
Design a scalable system to deduplicate and cluster billions of product listings across merchants into a unified catalog accessible to AI agents.
Automated data curation system using consensus from multiple LLM judges to identify when models should refuse low-confidence queries instead of generating unreliable responses.
Shopify's internal platform enables rapid deployment of web applications with minimal friction, transforming how engineers build and iterate on projects.
Airbnb's evolution from a single monolithic data system to a distributed, multi-tenant architecture supporting diverse product lines at scale.
Building and deploying a Slack-native AI agent required careful architectural decisions, infrastructure choices, and operational lessons learned.
Migrating inventory reservation system from Redis to MySQL using row-level locking and optimized indexing strategies to achieve production scale.
Fine-tuned open-source LLM as a tool-calling agent to convert natural language into workflow automations, reducing costs while maintaining accuracy through continuous learning.
Extending Karpathy's AutoML approach beyond training to systematically optimize 40+ business metrics at Shopify scale.
Design an AI-powered retail mirror system that processes real-time customer appearance data to generate personalized product recommendations using computer vision and backend services.
Shopify optimized GraphQL query execution by replacing depth-first traversal with breadth-first processing to reduce overhead and improve performance significantly.
Distributed testing platform using simulated robotic agents to validate e-commerce infrastructure and performance at scale.
Shopify built a generative recommendation system that models buyer journeys as sequences rather than isolated signals, enabling real-time personalization at commerce scale.
Shopify leverages SkyPilot framework to simplify multi-cloud GPU provisioning and management through unified configuration, eliminating vendor lock-in and streamlining development workflows.
Design of a universal commerce protocol enabling autonomous systems to execute transactions across multiple platforms and vendors at scale.
Distributed ML experimentation platform that reduces computation overhead, ensures reproducibility, and enables seamless experiment sharing across teams.
Engineering team built an interactive real-time globe visualization displaying live merchant sales data with gamified pinball mechanics for event engagement.
Engineers integrated fitness tracking sensors with real-time messaging and point-of-sale systems to create a treadmill that generates store credit for users.
Shopify's approach to preparing infrastructure for massive seasonal traffic spikes through simulation, chaos engineering, and bottleneck identification.
Shopify optimized product search by integrating machine learning models with C++ for high-performance query processing and ranking at scale.
Shopify uses AI agents to dynamically evolve product taxonomy at scale, moving beyond static classification systems to adapt categorization in real-time.
Teams transition mobile applications to React Native's improved architecture, addressing performance constraints and synchronization issues from the legacy system.
Shopify shares architectural patterns and testing strategies for deploying production AI assistants at scale with reliability.
Protocol extension enables AI agents to render interactive UI components for commerce, replacing text-only responses with visual product interfaces and transaction flows.
Major rewrite of a list virtualization library for React Native that improves scrolling performance and eliminates the need for pre-computed item dimensions.
Shopify uses multimodal large language models to intelligently organize and structure billions of product listings from diverse merchants into a unified global catalog system.
Shopify uses machine learning across their commerce platform to optimize merchant workflows and customer experiences at scale.