Building agent fleet architectures your CISO doesn't hate
summary
Gitpod moves from GCP Kubernetes to a bring your own cloud model that runs a tiny runner as a single ECS task (single‑digit dollars per month) on AWS EC2 backed by EBS, removing Kubernetes day‑two complexity. Customers keep source code and data in their accounts while metadata and audit logs live with the vendor, enabling auto‑suspend and resume for agent fleets. An agent offering launched two days ago runs inside the same dev environments with human‑equivalent access, producing full audit trails for regulated customers.
AI-written summary · about
chapters (11)
- 0:00Introduction and background
- 1:04Talk audience and purpose
- 3:12Architecture evolution overview
- 4:14Managed SAS and Kubernetes challenges
- 5:38Self-hosted model and drawbacks
- 7:03Substrate model and AWS focus
- 8:27Kubernetes abandonment and new architecture
- 9:50Current architecture and runner concept
- 11:36AI agent offering and benefits
- 12:39API-first approach and audit logs
- 13:45Closing remarks and Q&A
related talks
36mHow to Safely Deploy AI Agents that Write and Manage your IaC
06mThe End of Localhost: A Platform Team's Guide to Securing Agenti...
09mBuilding an Agent Runtime With Open Infrastructure
34mTaking Back Control: Building Cloud-Agnostic Container Platforms for Regulated Environments
49mPublic cloud in regulated industry, not just swipe of a cred
33mShipping Enterprise AI Agents with Amazon Bedrock AgentCore
14mTaming Agentic AI: Run Rogue Code Safely on Kubernetes with Agent Sandbox
25mWhat We Learned Securing AI Agents at Scale on Multi-Tenant Kubernetes Clusters