What is this?
This is a 4-night hands-on course that builds the data and sales analytics stack of a fictional B2B perfume distributor from scratch, using Databricks Free Edition, SQL, Python and Claude Code.
This is a 4-night hands-on course that builds the data and sales analytics stack of a fictional B2B perfume distributor from scratch, using Databricks Free Edition, SQL, Python and Claude Code.Is this a ready-made app or a course I work through?It is a project-based course: each night has its own self-contained folder with a README and numbered examples, and the whole build comes from pasting 12 prompts into Claude Code in sequence, each one ending in a deploy.
It is a project-based course: each night has its own self-contained folder with a README and numbered examples, and the whole build comes from pasting 12 prompts into Claude Code in sequence, each one ending in a deploy.How do I follow along?Start at aulas/aula-01-databricks-sql and go through the four night folders in order; the night-2 README has the full terminal setup (minimum versions, bundle init, skills, MCP, guard rails), and PROMPTS.md has every prompt ready to paste into Claude Code.
Start at aulas/aula-01-databricks-sql and go through the four night folders in order; the night-2 README has the full terminal setup (minimum versions, bundle init, skills, MCP, guard rails), and PROMPTS.md has every prompt ready to paste into Claude Code.What do I need before starting — a Databricks account, Claude Code?A Databricks Free Edition (serverless) workspace, Claude Code, plus SQL and Python; most commands also require the Databricks CLI with a --profile flag.
A Databricks Free Edition (serverless) workspace, Claude Code, plus SQL and Python; most commands also require the Databricks CLI with a --profile flag.Is the sample dataset already clean?No, the mess is intentional: CNPJs appear in three formats, dates in two, customers are duplicated, returns show up as negative quantities, and terminated sales reps still hold active portfolios — cleaning it is the actual content of night 2, not a bug to patch in the generator.
No, the mess is intentional: CNPJs appear in three formats, dates in two, customers are duplicated, returns show up as negative quantities, and terminated sales reps still hold active portfolios — cleaning it is the actual content of night 2, not a bug to patch in the generator.Is this course only in Portuguese?Yes, the README, prompts and all course material are written in Portuguese and modeled on a Brazilian business context (CNPJ, perfumaria árabe, revendedoras).
Yes, the README, prompts and all course material are written in Portuguese and modeled on a Brazilian business context (CNPJ, perfumaria árabe, revendedoras).What is the 'Rota do Perfume' story about?The course builds the data and sales stack for a fictional Arabic-perfume B2B distributor selling to perfumeries, pharmacies, resellers and e-commerces, with synthetic data generated using a fixed seed.
The course builds the data and sales stack for a fictional Arabic-perfume B2B distributor selling to perfumeries, pharmacies, resellers and e-commerces, with synthetic data generated using a fixed seed.What do the 12 prompts actually produce?Prompts 1-6 build the catalog, a 12-task pipeline, a dashboard and Genie; prompts 7-9 train a model registered in Unity Catalog; prompts 10-12 build an executive Genie, an app, and a callback flow.
Prompts 1-6 build the catalog, a 12-task pipeline, a dashboard and Genie; prompts 7-9 train a model registered in Unity Catalog; prompts 10-12 build an executive Genie, an app, and a callback flow.Does the data catalog already exist when I start?No, the lakehouserotaperfume catalog does not exist until it is created live during night 1 — if you already have a catalog under another name, you just swap the prefix in the queries.
No, the lakehouserotaperfume catalog does not exist until it is created live during night 1 — if you already have a catalog under another name, you just swap the prefix in the queries.What does the included dataset generator cover?scripts/material/gerardataset.py produces about 14 MB across ten tables (orders, order items, payments, products, stock, customers, visits, opportunities, portfolios, sales reps) spanning September 2024 to August 2026.
scripts/material/gerardataset.py produces about 14 MB across ten tables (orders, order items, payments, products, stock, customers, visits, opportunities, portfolios, sales reps) spanning September 2024 to August 2026.What are similar projects?hotdata-cli, airbyte, ai-engineering-from-scratch, WrenAI, AI-Engineering-Lab, potencializando-estudos-carreira-com-ia, turnstone, Gnos
hotdata-cli, airbyte, ai-engineering-from-scratch, WrenAI, AI-Engineering-Lab, potencializando-estudos-carreira-com-ia, turnstone, GnosWhich topics is it in?databricks, portuguese, data-pipeline, etl, data-engineering, course, sql, model-context-protocol, data-science, skills, anthropic, learning-resources
databricks, portuguese, data-pipeline, etl, data-engineering, course, sql, model-context-protocol, data-science, skills, anthropic, learning-resources