Summer 2026 edition — practitioner format
This workshop is built for people who already work with containers. Most participants use Docker, Kubernetes, or a managed variant (GKE, EKS, AKS) in their daily work, so the format is discussion- and challenge-driven rather than lecture- and walkthrough-driven:
- Discussions capture how containers are actually being adopted across campuses — what's working, what isn't, and what's next.
- Hands-on blocks are tracked. You pick the track that matches your experience and interests. Specs and stretch goals replace step-by-step instructions; full walkthroughs exist as a reference layer if you want them.
- The scenarios are real. Alongside a purpose-built app catalog, you'll build and connect LiteLLM and LibreChat — the gateway-plus-frontend stack a growing number of campuses deploy to broker and front AI access — locally first, then on GKE.
| # | Segment | Format | Time |
|---|---|---|---|
| — | Introductions & Goals | welcome | 20 min |
| 01 | Setting the Context | facilitated discussion | 20 min |
| 02 | Containers in Higher Ed | presentation + discussion | 30 min |
| — | Break | 10 min | |
| 03 | Hands-On 1: Build & Run | tracked lab | 45 min |
| 04 | Cloud & Orchestration | presentation + discussion | 25 min |
| — | Break | 10 min | |
| 05 | Hands-On 2: Kubernetes on GKE | tracked lab | 45 min |
| 06 | Wrap-Up & Takeaways | discussion | 15 min |
Scheduled time is 220 minutes; the remaining 20 minutes is deliberate buffer — hands-on blocks always run long.
Hands-On 1 offers two tracks; Hands-On 2 offers four. You do one track per block, chosen during the context discussion at the start of the workshop.
Hands-On 1 (local):
| Track | You will |
|---|---|
| A | Containerize the app catalog from a spec |
| B | Build the campus AI stack: a LiteLLM gateway with virtual keys and budgets, LibreChat in front, then wire them together — three milestones, each a complete stopping point |
Hands-On 2 (GKE):
| Track | You will |
|---|---|
| A | Deploy the app catalog to GKE, then scale it and fix what breaks |
| B | Deploy LiteLLM to GKE from the upstream chart |
| C | Deploy LibreChat to GKE from the official Helm chart |
| D | Architect a full campus AI stack (diagram + decision record) |
Rough guidance for choosing:
- Track A if you want a complete, verifiable build-to-cloud arc with the workshop app — or if containers are still newish to you. It has the most stretch goals, so daily Docker users won't run out of runway.
- Track B if you want the LiteLLM/LibreChat stack campuses are deploying for AI access. Its milestones feed directly into Hands-On 2's Tracks B, C, and D.
Every track has a Definition of Done, stretch goals, and a hint ladder (collapsed hints — open them only when stuck). Hands-On 1 Definitions of Done are checked by a verifier container you run locally — see verifier/.
You likely use AI coding tools daily; this workshop doesn't pretend otherwise.
- First pass: try the spec without AI assistance. The point is to find your own edges.
- Second pass and stretch goals: use whatever you use at work. Directing an assistant to produce a correct Dockerfile — and verifying what it produced — is a legitimate skill, and the verifier will tell you whether it got it right.
| Path | What it is |
|---|---|
| 00-prerequisites/ | What to install and verify before the workshop |
| 01-welcome-and-context/ | Discussion guide + notes template for the context session |
| 02-containers-in-context/ | Where containers fit in higher ed — good/poor fits, the adoption arc, and peer case studies |
| 03-hands-on-1/ | Hands-On 1 tracks (app catalog · campus AI stack) |
| 04-cloud-and-orchestration/ | Choosing an orchestrator, idea→prod, managed Kubernetes, and containerize-vs-buy |
| 05-hands-on-2/ | Hands-On 2 GKE tracks, campus requirements sheet, templates |
| 06-wrap-up/ | Wrap-up discussion guide |
| 07-cloud-cleanup/ | Tearing down every cloud resource created today |
| 08-reference/ | Full step-by-step app-catalog walkthrough (the "solution layer"), fundamentals primer, troubleshooting guides, and higher-ed container stories |
| gcp-deployment/ | Reference Kubernetes manifests for the app catalog |
| verifier/ | Source for the Definition-of-Done verifier container |
| instructor/ | Instructor runbook: provisioning, AI gateway, API keys, timing |
- Level: Intermediate. We assume working knowledge of git, Linux fundamentals, Docker, and at least passing familiarity with a public cloud and an orchestrator.
- Audience: Enterprise IT professionals, platform engineers, DevOps practitioners, and research computing facilitators in higher education.
- Delivery: Online via Zoom.
- Duration: 4 hours.
By the end of this workshop, you will be able to:
- Articulate where containers fit (and don't fit) in campus IT and research computing, drawing on adoption stories from peer institutions
- Build container images and multi-service Compose stacks against a spec, with security and operability requirements, and verify them objectively
- Deploy and evaluate the LiteLLM + LibreChat stack used by campuses to broker AI access — including virtual keys, budgets, and provider abstraction
- Wire a chat frontend to a governed gateway so provider credentials live in exactly one component, and prove it — the key-isolation pattern behind campus AI portals
- Adapt upstream Kubernetes artifacts (Helm charts, manifests) to institutional requirements, and produce a gap analysis for what the defaults don't cover
- Diagnose real orchestration failure modes: multi-replica session state, probe semantics, graceful shutdown, image/architecture mismatches