Workshop
Agentic Data Engineering in a Day: Level 1
Coding got faster. Data engineering did not. On the day, you're taking a real business request to a verified pipeline with Vibedata, then asking it to explain a changed number.
5×
The same team ships five times more with Vibedata.
Coding got faster. Data engineering did not. You still do the data engineering by hand, through the agent.
Cheapest model with Vibedata: 12 of 12. Most capable model alone: 5 of 12.
We expect you are doing most of these.
You, and the coding agent you already bought.
Before building
- You spec it every time. The next change starts blank.
- It asks only if told. Otherwise it guesses from code.
- What you taught it is in a chat log. Nothing carries.
- It writes fresh code to find what objects and jobs exist.
While building
- One shared pre-prod, synced by hand. Agents run one at a time.
- You watch it run so you can step in before it goes wrong.
- It writes fresh code to work out lineage and impact.
Review
- You read the whole diff and redo the checks. Design review by hand.
- It writes fresh code to tell you what passed and what changed.
After release
- A business user finds the wrong number first, or you check by hand.
- Health is ad-hoc queries somebody writes and runs each time.
- Each fix stays a one-off. The same question returns next month.
Your time · five times slower than possible.Your tokens · fresh code, not a tool.
Each one costs you time or tokens. On the day, we're showing you how to leave them all behind.
What you leave with
Instead of: You spec it every time.A spec, sharpened by pushback.
You're turning a real business request into an approved Intent with success criteria and source assumptions.
Instead of: What you taught it is in a chat log.A pipeline built your way.
You're adding one house rule. Vibedata is building the ingestion and processing pipeline in its own branch.
Instead of: You read the whole diff and redo the checks.A defect caught before it ships.
You're finding the seeded defect against golden data, fixing the spec and reaching a certified verdict.
Instead of: A business user finds the wrong number first, or you check by hand.A changed number, explained.
Vibedata is tracing why a shipped number moved and raising an issue.
Who it is for
25 Nov 2026 · 10:00–16:00 SGT, doors 09:15
Mapletree Business City, Singapore
For data engineers who already use AI to write SQL
Bring a laptop · your own Domain in a hosted sandboxYour own platform is the Level 2 day.
Agenda
Timings are indicative and will move with the day.
- 09:15–10:00 · Registration and sandbox sign-in
- 10:00–10:30 · Intro
- 10:30–11:25 · Exercise 1a: From business request to spec
- 11:25–11:35 · Break
- 11:35–12:45 · Exercise 1b: Design, plan, build
- 12:45–13:30 · Lunch
- 13:30–14:15 · Exercise 2: Catch a defect before it ships
- 14:15–15:00 · Exercise 3: Explain a changed number
- 15:00–16:00 · Bringing it all together and closing