Data Mozart
live workshop - September 7, 2026 - 5pm Central Europe
4-hour live workshop - Fabric implementation - portable agentic patterns

AI-Native Data Engineering with Microsoft Fabric

Build a metadata-driven Fabric ingestion and medallion scaffold with coding agents - then master the skill that travels across platforms: specifying, bounding, and validating generated data engineering work.

September 7, 2026 - 5pm Central Europe / 11am EDT / 8am PT - live online - 4 hours
Reserve your seat See the agenda
Taught by Nikola Ilic (Data Mozart) - €179FOUNDING COHORT
fabric - agent session
$ agent "scaffold a medallion workspace for card payments"
> Planning: workspace, 3 lakehouses, 4 notebooks, 1 pipeline
$ fab create ws_payments.Workspace
$ fab create ws_payments.Workspace/lh_bronze.Lakehouse
$ fab create ws_payments.Workspace/lh_silver.Lakehouse
$ fab import nb_transform.Notebook -i ./src
> Generated: bronze-to-silver transform, 214 lines
your job: find the join that silently drops rows.
# that's what this workshop teaches.
$
why this workshop exists

Data engineering is becoming less about typing and more about judgment.

Microsoft is shipping assistant-driven workflows into Fabric: Copilot in the web experience, developer agents in VS Code, and a CLI that can turn workspace changes into repeatable commands. Databricks, Snowflake, and the rest of the modern data stack are moving in the same direction: more generated code, more resources controlled by automation, and more responsibility on the engineer to verify what happened.

Here's what the demos skip: generated code is confidently wrong in ways that only show up downstream. A plausible join that silently drops rows. A schema assumption nobody made. A deprecated pattern dressed up as best practice.

The engineers who win in this shift aren't the fastest typists. They're the ones who can specify precisely, automate aggressively, and validate ruthlessly. We implement the workflow in Fabric, but the judgment muscle transfers: guardrails, metadata contracts, validation evidence, deployment checkpoints, cost boundaries, and security review.

what you'll build, with your own hands

One focused build. Automation where it actually helps.

The core lab is a metadata-driven Fabric ingestion and medallion scaffold. You leave with the working Fabric solution, plus validation patterns, agent guardrails, and a prompt/scaffold library you can adapt to other modern data platforms.

scaffold

A medallion workspace from a spec

Lakehouses, notebooks, and pipelines scaffolded by a coding agent driving the Fabric CLI - from an agent skill you author, with guardrails. Zero portal clicking.

framework - centerpiece

A metadata-driven ingestion framework

One configurable pattern instead of N pipelines - config tables, parameterized notebooks, dynamic sources - generated with an agent and audited where the blast radius is largest.

transform

Generated pipelines you've verified

Bronze-to-Silver-to-Gold transformations drafted with coding assistance - then audited by you, with the planted failure modes found and fixed.

enrich

LLM transforms inside the notebook

An LLM called straight from a Fabric notebook with your own API key - classification, extraction, summarization - as proper pipeline steps. Designed for the trial path, no paid Fabric Copilot capacity needed.

automate + ship

Generated tests, docs, and deployment pattern

The unloved work produced automatically and matched to a gold-standard exemplar you provide - with a CLI deployment pattern included so you can promote the solution after the workshop.

transfer

A platform translation map

What stays the same when you move from Fabric to Databricks, Snowflake, or another stack - and what must be rebuilt because every platform has its own CLI, security model, runtime, and deployment surface.

who this is for

Built for working data people, not prompt tourists.

This is a practical engineering workshop. You do not need to be a Fabric expert already, but you should be comfortable reading code and thinking carefully about data quality.

Good fit

  • Data engineers, analytics engineers, architects, and senior BI developers moving into Fabric.
  • Teams that want coding agents to speed up scaffolding, tests, docs, and repetitive pipeline work.
  • Practitioners who care about validation, cost, security, and production readiness as much as generation speed.
  • Fabric-first teams, plus Databricks/Snowflake teams that want the transferable workflow and can adapt the platform layer.

Not a good fit

  • If you want a no-code tour of Fabric screens, this is too engineering-heavy.
  • If you expect a finished production platform after four hours, that would be dishonest.
  • If you cannot read Python or SQL at all, the review exercises will feel too fast.
  • If you want unsupervised agents changing production tenants, this workshop will annoy you in a healthy way.
Fabric lab, portable workflow

The workflow transfers. The platform layer changes.

We implement the build in Microsoft Fabric because concrete beats abstract. But the professional habit you are practicing is broader than one vendor.

What transfers

Bounded tasks, repo instructions, metadata contracts, scaffold prompts, validation checklists, generated test/docs patterns, LLM enrichment boundaries, cost review, and human deployment checkpoints.

What changes

CLI commands, resource definitions, identity and permissions, runtime behavior, deployment artifacts, observability, and cost model. Fabric users get the direct implementation; other platform teams get the map.

Why that matters

The dangerous part is rarely the button. It is the unchecked assumption: an inner join, inferred schema, overwrite mode, missing source behavior, or quietly duplicated fact table. Those mistakes follow you across platforms.

the four hours

Agenda

Every session moves the same Fabric build forward. The center of gravity is the ingestion framework and the validation habits around it; along the way, we call out which parts are Fabric-specific and which parts are portable engineering patterns.

0:00 - 0:20

What "AI-native" actually means

The evolving role, the hype, and the three surfaces: Copilot in Fabric web for exploration, VS Code for developer workflows, the CLI for automation - and how to match the interaction model to the task. Plus the mental model the whole day runs on: agents in a context-bounded box with a well-defined task.

0:20 - 1:05

Agentic scaffolding and the skill that builds your workspace

Fabric items as code instead of clicks. You author an agent skill with real guardrails - explicit targets, pause-for-review checkpoints, "present the plan and stop", "you may not commit" - then drive the agent to scaffold the medallion workspace from that specification.

1:05 - 1:15
break
1:15 - 2:10

Building the metadata-driven ingestion framework CENTERPIECE

One configurable pattern instead of N hand-written pipelines: config tables, parameterized notebooks, dynamic sources - generated with an agent. Then the audit that matters most, because a confidently-wrong config assumption breaks every table the framework drives at once. Reading generated PySpark critically is a different skill from writing it, and this is where you learn it.

2:10 - 2:20
break
2:20 - 3:10

Transformations, LLM enrichment, and automating the unloved work

Bronze-to-Silver-to-Gold driven phase by phase using the gold-standard exemplar pattern, so generated code matches your bar, not the model's guess. LLM-powered transforms called from the notebook with your own API key (classification, extraction, summarization), designed for the trial path. Then generating tests, quality checks, and documentation from what you built, with deterministic tools for repeatable agent behavior.

3:10 - 3:20
break
3:20 - 4:00

Deploy, trust, cost, and the production checklist

A CLI-driven deployment pattern, then the honest part no demo shows: what model features cost in capacity units, the security implications of agents on your tenant, an honest capability map of every surface, and what to never automate. Closes with where Data Agents fit on the consumption side of Fabric, how the same workflow maps to Databricks/Snowflake-style teams, and open Q&A.

the question everyone asks

"Do I need a paid Fabric capacity?"

No - and I'm not going to pretend the paid Copilot features don't exist, or quietly make you buy them. The workshop is designed around the licensing reality.

  • The hands-on labs are designed for the free 60-day Fabric trial capacity. Scaffolding, the metadata framework, transformations, LLM enrichment (via your own API key), tests, docs, and the deployment pattern use the agent-plus-CLI workflow, not paid Fabric Copilot capacity. The setup kit has you verify your tenant before workshop day.
  • Only the native Fabric Copilot segments run as guided demo-along, since Copilot and Fabric's built-in model functions require a paid F capacity. If your org has one you can follow live; if not, you miss nothing hands-on.
  • You do not need to buy Microsoft capacity to get full value. Bring your own LLM API key for the enrichment lab; expected usage is tiny, but it may create a small provider charge. Curious about native Copilot? Check current Azure pricing for F2 or above, because price varies by region and currency.
founding cohort pricing

First run. Lowest price it will ever be.

This is the founding cohort - the price goes up once this run's results and testimonials are in. In exchange, I'll ask you for honest feedback.

Standard

€179
open registration
  • The live 4-hour workshop, Q&A throughout
  • Full recording, yours to keep
  • Scaffold repo + prompt library, with 12 months of updates
  • Pre-workshop setup kit and smoke test
  • Agent-readiness production checklist
Reserve Standard - €179
15 SEATS ONLY

Pro

€449
limited - access does not scale
  • Everything in Standard
  • Live office hours, 2-3 weeks after - once you have hit real questions on your own tenant
  • 30-minute 1:1 call to review your setup or use case, scheduled within 30 days
  • Priority Q&A during the live session
Reserve Pro - €449

Paying through your company? Most attendees do. You'll get a proper VAT invoice and a ready-made justification one-pager for your manager. Team pack: 5 seats for €795. Private team workshop: from €6,500. Email for team invoices and private dates.

Minimum cohort size: the live workshop will run if at least 10 participants are registered. If the minimum is not reached, you will be offered a full refund or the option to transfer your seat to the next scheduled cohort.

Nikola Ilic - Data Mozart
your instructor

I built this workflow before Microsoft started marketing it.

I'm Nikola Ilic, better known as Data Mozart. Microsoft Data Platform MVP, Principal Data Architect at iLink Digital, book author, and creator of Fabric and Power BI training that has reached thousands of professionals through Pluralsight, O'Reilly, conferences, and the Data Mozart blog and YouTube channel.

The agentic CLI workflow at the center of this workshop comes from my own consulting projects - the field version, not the keynote version. We implement it in Fabric because that gives us a concrete build, but the real lesson is broader: how to direct coding agents safely inside a data engineering workflow. What works, what breaks, what it costs. No vendor pitch, because no vendor is paying for this.

Microsoft Data Platform MVP - author of "Fundamentals of Microsoft Fabric" and "Analytics Engineering with Microsoft Fabric and Power BI" - Pluralsight + O'Reilly instructor - FabCon speaker
7d

The guarantee: attend the full workshop, and if you don't feel it was worth the price, email me within 7 days for a full refund. No awkward forms, no hard feelings.

questions, answered straight

FAQ

Do I need a paid Fabric capacity?
No. The hands-on labs are designed for the free 60-day Fabric trial capacity, including the LLM enrichment, which uses your own API key from a notebook rather than Fabric's paid model functions. Only the native Fabric Copilot demo segments need a paid F capacity, and you miss nothing essential without one. The setup kit includes a pre-flight check because some corporate tenants disable trials.
Do I need a Fabric tenant at all?
You need a work or trial account where you can activate a Fabric trial. Some locked-down corporate tenants block trials, so the setup kit walks you through the check days ahead and suggests a personal or dev tenant fallback if needed.
Is this useful if my company also uses Databricks or Snowflake?
Yes, if you want the agentic engineering workflow, not just Fabric button training. The hands-on implementation is in Microsoft Fabric, so Fabric users get the most direct value. But the core patterns translate: repo-based agent guardrails, metadata-driven ingestion, transformation validation, LLM enrichment boundaries, generated tests/docs, deployment checkpoints, and production trust review. The platform-specific layer - CLI commands, security model, runtime behavior, and deployment artifacts - must still be adapted.
Which coding agent do we use?
The live path covers GitHub Copilot in VS Code and Claude Code. The workflow is intentionally agent-portable, but those are the two stacks the setup kit and demos focus on.
How much coding experience do I need?
You should be comfortable reading Python and SQL and editing what's there. The whole point is that the assistant does much of the typing - your job is specifying and validating, and that's exactly what we practice.
I can't make the live date. Should I still register?
You get the recording, the repo, and 12 months of updates either way. But the live Q&A and the "find what the agent broke" exercises are better live.
Is there a minimum number of participants?
Yes. The workshop runs if at least 10 participants are registered. If the minimum is not reached, you will be offered a full refund or the option to transfer your seat to the next scheduled cohort.
Will this be outdated in six months?
The specific buttons will move - that's Fabric. The workflow (specify, scaffold, validate, automate) won't, and your seat includes 12 months of repo and prompt-library updates as the tooling evolves.
Can I buy seats for my team?
Yes. The public team pack is 5 seats for €795 with a single invoice. If you want this delivered privately for your team, private workshops start at €6,500 and can be tailored around your tenant, standards, and governance questions.
Why is this cheaper than comparable workshops?
Founding cohort. Future runs will cost more. In exchange, I'll ask for your honest feedback and, if you're happy, a testimonial.
SEPTEMBER 7, 2026 - 5PM CENTRAL EUROPE / 11AM EDT / 8AM PT - FOUNDING COHORT

The engineers who thrive won't be the fastest typists.

They'll be the ones who can direct coding agents precisely and catch their mistakes confidently. Four hours from now, you can have the Fabric scaffold, the portable validation habits, and a clear-eyed view of what these tools really do.

Reserve your seat - €179