Dataiku: Flow, Recipes & Datasets

Dataiku (branded Dataiku DSS -- Data Science Studio) is an end-to-end data science and machine learning platform combining visual, no-code tooling with code-based work (Python/R/SQL) in one collaborative environment -- letting both technical and less-technical team members contribute to the same project.

Projects, Flows & Datasets

Project: "Customer Churn Analysis"
  |
  +-- Flow (the visual pipeline diagram)
        |
        +-- Dataset: raw_customers.csv
        |     |
        |     +-- Recipe: Prepare (visual, no-code cleaning)
        |           |
        |           +-- Dataset: customers_cleaned
        |                 |
        |                 +-- Recipe: Train (visual ML)
        |                       |
        |                       +-- Model: churn_predictor

# A dataset can reference: a file (CSV, Excel), a database table
# (SQL connection), cloud storage (S3, GCS), or another recipe's
# output -- a unified interface regardless of where data lives

Recipes: Visual and Code-Based

A visual 'Prepare' recipe lets a business analyst clean up messy column names and handle missing values through point-and-click steps -- no coding required, but leveraging their deep domain knowledge of what actually matters for the business. A Python/R/SQL recipe handles logic too complex to express visually. Both coexist in the same Flow.

Visual Machine Learning

A guided, largely no-code interface for building and evaluating models -- select a target variable, choose from common algorithms, and Dataiku handles train/test splitting, preprocessing, and evaluation metrics automatically. A data scientist can still drop into custom code when the visual approach isn't sufficient.

Choosing Visual vs. Code

Visual recipes fit common, straightforward transformations (renaming columns, filtering, type conversion) that are readable to a broad audience. Code becomes appropriate as logic grows genuinely complex or benefits from version control/testing -- most real projects mix both based on each step's actual needs.