Practitioner collateral from KodeKloud webinars. Designed to be used the week after you watch them, not filed away.
The four NFCU sessions below form a sequential, hands-on live-demo series on ML model deployment for DevOps engineers in regulated financial services — each session's outputs feed the next. Every session directory is self-contained and has its own README with role-based starting points.
Collateral from "Beyond Copilots: How Agentic AI is Rewriting the DevOps Playbook" (April 2026).
Templates, checklists, and a structured 90-day path for running your first agentic AI pilot in a DevOps workflow. Includes:
- A governance-first 90-day playbook for CI build failure analysis
- A team readiness assessment scorecard
- Agent boundary design templates (AGENTS.md and permission matrix)
- An eval framework with 10 starter scenarios
- A tool landscape reference organized by the 4-layer stack
- An annotated external reading list
See the Agentic DevOps README for role-based starting points.
Runnable collateral from Session 1 of a four-part live-demo series on ML model deployment for DevOps engineers (June 2026).
A complete, traceable deployment pipeline built around one question: can you trace any prediction back to its training data in under five minutes? Includes:
- A deterministic sample-model training script and an immutable artifact contract
- A validation gate that rejects mutable references, plus container build/scan/sign and audit-trail tooling
- Reusable Terraform modules for VPC-isolated, KMS-encrypted SageMaker endpoints across dev/staging/production
- A manually-applied lab platform sandbox (OIDC, IAM, buckets, ECR, KMS)
- Three OIDC-only GitHub Actions workflows with SHA-pinned third-party actions
- A lab guide, FAQ, and an annotated reading list
See the NFCU Session 1 README for role-based starting points.
Runnable lab artifacts for Session 2 (June 4, 2026) — running a challenger model in production shadow alongside the champion, with zero customer impact. Includes:
- A shadow-mirror Lambda that mirrors live traffic to the challenger and a comparison Lambda that scores agreement
- A traffic generator, CloudWatch dashboard, and a reusable
audit_trailTerraform module (consumed by Session 3) - An OpenSpec change proposal as the source of truth, plus unit/integration tests
See the NFCU Session 2 README for the full run guide.
Collateral and lab environment for Session 3 (June 16, 2026) — monitoring, drift detection, and observability for deployed models, built to catch the model that returns HTTP 200 while quietly predicting wrong. Includes:
- Five Lambda functions: PSI drift-detector, drift-simulator, Evidently runner, NannyML runner, and an incident-simulator
- A CloudWatch dashboard and alarms, per-attendee Terraform, and a 5-phase runbook set
- Unit + moto-mocked integration tests and dry-run result docs
See the NFCU Session 3 README for the self-contained run guide.
Collateral for Session 4 (June 18, 2026) — moving the same workload off SageMaker onto Kubernetes-native serving with KServe. Runs locally on kind with no cloud spend, or on your own EKS via the included Terraform. Includes:
- A scale-to-zero
InferenceService, a CPU-only LLM, per-model cost attribution, and a canary rollout with rollback - A local kind cluster with add-ons, EKS Terraform, and CI
- Four hands-on labs in the attendee guide plus a full-session rehearsal script
See the NFCU Session 4 README for quickstart and labs.
Clone, fork, or download the files you need. The templates include filled examples to show the format — replace the example content with your own infrastructure, tools, and failure patterns.
Documentation content is licensed under CC BY 4.0. Code snippets within the documentation are licensed under Apache 2.0.