Erik Bernhardsson, former Spotify engineer and CTO, introduces Modal, a cloud compute platform designed specifically for data teams. Modal lets developers run Python functions in the cloud in under a second using a simple decorator syntax, without writing YAML or managing Kubernetes. It handles containerization, GPU access, scheduling, and scaling automatically. Under the hood, Modal uses custom-built container runners, image builders, and a FUSE-based filesystem written in Rust to achieve fast cold starts and efficient image management. Use cases include ML model inference (Stable Diffusion, Whisper), data pipelines, cron jobs, and web scraping. The product was in waitlist-only mode at time of writing.
Table of contents
ContextData as its own disciplineLet’s make data teams more productive!Let’s build a new runtimeWhat are some things you can build using Modal?How does it work?Sort: