Just in Time Infrastructure
summary
A data science platform is built by a team of five that ties application code and infrastructure code together, using a Python-based Particle Cloud Framework (open-sourced by Capital One) plus Terraform/CloudFormation for templates, and Cloud Custodian for governance. The system declaratively spins up full stacks on demand — demonstrated by provisioning 60–70 GPU clusters for a tutorial — and supports concurrent dev environments, CI/CD that deploys infra and app together, cost-saving teardown during business hours, and automated rollback and self-healing via lambdas.
AI-written summary · about
chapters (10)
- 0:00Start
- 1:32Dynamic Infrastructure
- 4:10What Is Just-in-Time Infrastructure
- 6:15Cloud Resource Management Framework
- 6:46Particle Cloud Framework
- 7:35Core Benefits
- 8:04Application Isolation
- 8:42Cost Management
- 18:14How Do You Find that It Scales To Like Larger Teams with You Know Multiple Branches Going On at Once
- 20:47Who Maintains the Infrastructures Code
related talks
27mAdvanced Multipurpose Infrastructure: A Self-Managed Cloud Environment with HashiCorp Terraform
27mApplication-Driven Infrastructure
30mPython-friendly cloud infrastructure management
01h 01mTaming the Modern Data Center
22mTerraform Hybrid: Unifying Your Physical and Cloud Datacenters
27mCase Study: Re-Thinking Our Infrastructure Tooling
22mEnabling Infrastructure as Code
55mSelf-service infrastructure is no longer a dream