Emma Forman Ling: Sponsor Lightning Talk: "Scaling Infrastructure With Confidence in the Age of AI"

Hi everyone. Um my name is Emma and my pronouns are they/them and I am the tech lead on Convex's infrastructure team and today I'm super excited to talk to you all about how we think about scaling infrastructure with confidence in the age of AI. So first, what do I mean about scale? Um in the past 4 years I've worked at Convex, we've gone from

10 to millions of deployments and some of those deployments have thousands of queries per second, but our team has not grown at that pace. Uh we've gone from 7 to 14 engineers. So, how are we managing to do this? We live in this abundant world, right? We're probably just sending AI agents to go build all of these systems to scale. Um but that's not quite what's happening

here. Uh we it's a little bit more nuanced than that because Convex is a database company. Correctness matters a lot. Reliability matters. If we ship a bug that makes uh Convex inconsistent, that means our customers' data will be corrupted in ways that we cannot fix. So, we have to use a lot of care when we use AI agents and um what what we found through

experimenting a lot with like trying to get AI agents to go write systems from scratch and like trying to set abstractions ourselves and then use those um to as guardrails for AI agents is that agents really like guardrails and good abstractions. Good abstractions keep the code simple and testable and that is what makes it possible to leverage AI agents effectively

to give you more confidence in your code being production ready rather than less confidence. Um so I'm going to dive deep into a project that we worked on. Um but first a bit of background. Oops. Um so how does Convex work? Um basically, our users write TypeScript functions that execute as acid transactions inside the database. The database itself and the back-end systems are