Architecture at QCon San Francisco 2025

Modern AI platforms don’t have to choose between deterministic precision and probabilistic exploration—they need both.

These days it is possible to achieve fairly good performance on cloud provisioned systems. We discuss the design of a high performance, strongly consistent system which maintains constant service in the face of regular updates to core logic.

What if everything you know about building distributed systems is backwards?

Cloud systems are undergoing an architectural shift. Traditional shared-nothing designs struggle to deliver the elasticity, availability, and operational simplicity that the cloud demands.

In this talk, Matthew will describe lessons learned from over 20+ years of building scalable, secure and stable infrastructure platforms for software in financial services (electronic trading, credit card processing etc.), the talk is relevant to anyone building platforms for mission-critic

What does it take to safely delete data at Netflix scale? In large-scale systems, data deletion cuts across infrastructure, reliability, and performance complexities.

Details coming soon.

How does Netflix maintain a seamless viewing experience for millions of users, especially during traffic spikes or when backend datastores are overloaded? Autoscaling can help during traffic spikes, but it costs money, takes a few minutes to kick in, and capacity may not always be available.

Front-end has long been about reactivity frameworks and client-side state management. However, the alpha in these is receding.

One does not simply migrate a monolith. Imagine a team working on a monolith-to-microservices migration of a healthcare portal. A foundational first step - migrating to a commercial identity provider - takes 9 months, only to bring the entire portal crashing down on release day.

How do you deliver relevant and personalized recommendations to nearly a billion professionals—instantly, reliably, and at scale? At LinkedIn, the answer has been a multi-year journey of architectural reinvention.

Stripe processes over $1 trillion in payments annually with industry-leading reliability, powered by its custom-built document database, DocDB, built on top of open source MongoDB. Stripe's DocDB serves over five million queries per second from Stripe’s product applications.

Uber Eats has many surfaces developed using native-first design. Historically these were built on the Android and iOS stacks. To accelerate development and enable rapid iteration and experimentation, while preserving the native-first design, a webview-powered stack was developed.

At Netflix, certain use cases demand the rapid transfer of massive datasets—such as 50 TB—from offline to online systems. Doing this efficiently, without disrupting applications interacting with our online systems, presents a significant challenge.

AI inference is expensive, but it doesn’t have to be. In this talk, we’ll break down how to systematically drive down the cost per token across different types of AI workloads.

Netflix runs on a complex multi-layer cloud architecture made up of thousands of services, caches, and databases. As hardware options, workload patterns, cost dynamics and the Netflix products evolve, the cost-optimal hardware and configuration for running our services is constantly changing.

Ever wondered what happens when a bunch of performance-obsessed developers decide their blazing-fast database isn't quite blazing-fast enough?

SS&C Technologies runs 47 trillion dollars of assets on our global private cloud. We have the primitives for infrastructure as well as platforms as a service like Kubernetes, Kafka, NiFi, Databases, etc.

As AI applications demand faster and more intelligent data access, traditional caching strategies are hitting performance and reliability limits.

As companies grow, so does the complexity of keeping distributed systems in sync. At DoorDash, we tackled this challenge while building a high-throughput, domain-oriented data platform for capturing changes across hundreds of services.