Knowledge HubbyDataminded

AI Engineering Capstone

Improve one agent in a dark factory for data products: a line of agents that builds a data product from a request, with no human in the loop.

AI
intermediate
8h
View exercises

Overview

A dark factory is a production line that runs without people. Here the line builds data products: a chain of agents takes a stakeholder's request to tested dbt models and a dashboard. Each team owns one agent in that line. You measure it on a benchmark, improve it through its context, and at the end of the day all agents run together on a request nobody has seen.

What you'll cover

  • ▸The dark factory: a line of agents that builds data products without a human in the loop
  • ▸Contracts: the fixed file formats agents use to hand work to each other
  • ▸Executable specifications: requirements that carry their own acceptance checks
  • ▸Hold-out sets: unseen benchmark cases that expose overfitting

What you'll be able to do

  • ✓Measure a baseline agent on a benchmark and extend it with cases of your own
  • ✓Improve an agent through its context and prove each change with numbers
  • ✓Plug your agent into a line with other teams' agents

Tags

AI AcademyAgentic AIEvaluation

Prerequisites

  • The earlier modules of the AI Engineering Bootcamp, or equivalent experience
  • Working knowledge of Python and SQL

Technologies

Related items

Additional material

No additional material.