Block 1 / Topic 1.1 — Kickoff + framing

1.1.1

Thesis

3 min Lecture

Slide 1 / 2 · Title

AI Engineering: From Prompt Architecture to Production Infrastructure

"Stop building demos. Start shipping AI systems that actually work."

Martin Rojas + Samuel Obukwelu · VueConf US 2026 · All day

Slide 2 / 2 · The gap

Most developers can…

Call an API ✓
Write a prompt ✓
Build AI systems that consistently work in production ✗

Today closes the third gap.

Speaker notes

Open hard with the thesis: "Stop building demos. Start shipping AI systems that actually work."

Then the gap: "Most developers in this room can call an API. Most can write a prompt. Very few have built AI systems that consistently deliver production-quality results in enterprise environments — the kind that survive compliance review, handle multi-LLM failover, and don't regress silently when a model version changes. Today closes that gap."

Anchor: "Two halves. Morning: engineer prompts that work — structural patterns and constraint systems that eliminate the prompt lottery. Afternoon: production infrastructure — gateway architecture, governance, the concerns tutorials skip. Both halves teach engineering, not creative writing."

1.1.2

Four altitudes

4 min Lecture

Slide 1 / 2 · The frame

The day, in one frame

INFRASTRUCTURE  (Block 4)   ← AI gateway, routing, governance
WORKFLOW        (Block 3)   ← composing skills, hooks, multi-CLI
SKILL           (Block 2)   ← promote prompt → versioned skill + evals
PROMPT          (Block 1)   ← structural patterns, constraint engineering

Constraint engineering is the discipline. The artifact splits.

Slide 2 / 2 · Schedule

Day shape

Time	Block	Owner
`9:00 – 10:30`	1. Foundations	Martin
`10:30 – 10:45`	Break	—
`10:45 – 12:00`	2. Build & Eval a Skill	Martin
`12:00 – 1:00`	Lunch	—
`1:00 – 2:30`	3. Vibe Coding → Agentic Engineering	Sam
`2:30 – 2:45`	Break	—
`2:45 – 5:00`	4. AI Gateway (Production)	Martin

Speaker notes

Slide 1 — name the four altitudes: "The day has four altitudes. Prompt at the bottom — what most workshops focus on. Skill above it — the same prompt promoted to something repeatable and tested. Workflow above that — how a working engineer composes skills, hooks, and tools every day; that's Sam's block. Infrastructure on top — production gateway, governance, the concerns tutorials skip. The discipline is the same at each altitude: constraint engineering. The artifact splits."

Slide 2 — walk the schedule once: "Four blocks. Two breaks, one lunch. Sam takes Block 3 after lunch — his block is the bridge between 'you can build a skill' (Block 2) and 'you can deploy fleets of skills' (Block 4)."

Anchor: "By the end of today you've worked at all four altitudes on the same example. That's the through-line."

1.1.3

Intros + tool poll + housekeeping

3 min Lecture

Slide 1 / 3 · Co-presenters

Who's teaching

Martin Rojas — Blocks 1, 2, 4
Samuel Obukwelu — Block 3 (Vibe → Agentic)
2 voices, 4 altitudes, 1 day

Slide 2 / 3 · Tool poll

Which CLI did you bring?

Claude Code
Codex CLI
Cursor / Gemini CLI
Other / none yet

Calibration data for Topic 1.3.

Slide 3 / 3 · Housekeeping

Before we start

The workshop repo: git clone …ai-engineering-workshop (see Prerequisites)
skill-creator plugin should be installed (we'll verify at end of Block 1)
Need restroom / coffee? Two breaks: 10:30 and 2:30
Stragglers: back-of-room helper available all day

Speaker notes

Slide 1 — short co-presenter intros: "I'm Martin. Sam's here too — he's leading Block 3 after lunch. Two voices, four altitudes, one day." Sam stands or waves.

Slide 2 — the tool poll: "Show of hands. Who brought Claude Code? [count]. Codex CLI? [count]. Cursor or Gemini? [count]. None / not sure yet? [count]." This is calibration data for Topic 1.3 — if the room is mostly Claude Code, you can lean in; if it's split, you'll spend more time on the cross-CLI mapping.

Slide 3 — housekeeping: "Repo URL is in the Prerequisites doc, also on the screen. The skill-creator plugin should be installed — we'll verify together at the end of Block 1. Breaks at 10:30 and 2:30, lunch at noon."

Close: "Let's get into it."