The Infra Pod Future of File Storage for AI (Chat with Hunter, CEO of Archill)
Sep 8, 2025
Hunter Leath, CEO of Archil, is on a mission to redefine cloud storage, particularly for AI applications. He discusses the limitations of current storage solutions like S3 and the revolutionary potential of using SSDs and custom protocols. Hunter highlights the complexities of managing unstructured data and why trust in infrastructure is critical. The conversation also ventures into future possibilities for file storage as well as practical applications ranging from serverless Jupyter notebooks to the demands of continuous integration and delivery.
AI Snips
Chapters
Transcript
Episode notes
Infinite Storage Attached To Servers
- Cloud object storage hasn't changed how files are used; servers still need local disks and manual syncing.
- Archil aims to expose infinite, synchronized storage to servers with local-like semantics.
Unstructured Data Drives The Problem
- "Unstructured data" means images, PDFs, and files used by ML and RAG workflows stored in S3 but downloaded to servers to use.
- These workloads drive the need for a POSIX-like, performant shared file layer.
SSD SAN With Custom Protocol
- Previous attempts exposed object storage but either lacked POSIX compatibility or were too slow.
- Archil builds an SSD-backed SAN and custom protocol to deliver block-like, local speeds with file semantics.