Event details
Most conversations about agentic AI sandboxes start and end with isolation: Can the agent run code safely? How fast does the container boot? What happens when an autonomous agent begins to engage in real, complex software engineering or data science tasks?
An agent doesn't just execute a single command and vanish. It clones massive codebases, installs deep Node modules or Python virtual environments, downloads browser binaries for testing, compiles artifacts, and writes gigabytes of logs and traces. The files are not incidental—they are the working memory of the task. If a sandbox relies entirely on a small local boot disk, the disk fills up, builds corrupt, and the agent fails.
Modern AI systems need to keep things separate: a reliable boot disk for the operating system and a large, long-lasting file system for the actual tasks and durable memory.
In this month's office hours, JuiceFS and Celesto AI team up to explore the engineering philosophy behind building high-performance, durable infrastructure for AI agents. We will break down how Celesto AI designed its agent sandboxes using CelestoFS (powered by JuiceFS under the hood) to seamlessly turn affordable object storage into highly performant, POSIX-compliant workspace storage. Finally, we will outline concrete scenarios showing how developers can leverage these sandbox environments for end-to-end AI software engineering, systems testing, and data analysis tasks.
Agenda
- Introduction to Our Guest: Celesto AI
- Agentic AI Storage Problem: The Boot Disk Should Not Be the Workspace
- The Sandbox Architecture: How Celesto AI Solves the Problem
- Under the Hood: How JuiceFS Bridges the Gap Between Object Storage and Local File Systems
- Practical Scenarios: Running Heavy-Duty Coding, Testing, and Data Agents without Disk Failures
- Live Q&A: Ask the JuiceFS & Celesto AI Teams Anything
Speakers
- Aniket Maurya, Founder @Celesto AI
- Joe Zhou, Developer Advocate @JuiceFS
JuiceFS Office Hours
The JuiceFS Office Hours is a recurring online event held on the fourth Thursday of each month. It serves as a dedicated platform for JuiceFS users and community members to:
- Share best practices and community experiences around AI/ML workloads.
- Receive immediate responses to their questions from the JuiceFS team.
- Learn about the latest product updates and feature releases.
