PROJECT

Atlas DataFlow

Interactive platform for publishing dataset studies and predictive experiences with contract-governed releases.

Transforms classification, regression, and forecasting studies into a public catalog with metrics, visualizations, documentation, and interactive inference where applicable

Status: Live

Open application Repository

What the project offers

In Atlas, each published dataset receives its own detail page. Visitors can explore the problem being studied, understand the target, review performance metrics, navigate analytical visualizations, and read the technical documentation behind the study. When supported by the study, the project also provides an interactive inference experience.

Atlas DataFlow dataset overview showing performance metrics, target distribution, feature importance, and confusion matrix.
Dataset overview
Atlas DataFlow interactive inference form showing customer inputs and a churn prediction result.
Interactive inference
Atlas DataFlow documentation view showing the Dry Bean study summary and dataset overview.
Technical documentation

How it works

Behind the public interface, Atlas connects integration notebooks, validation routines, contracts, evidence, candidates, and immutable releases. This structure keeps each study traceable and helps align public content, model behavior, metrics, and runtime without depending on the original notebooks in production.

Atlas DataFlow admin dashboard showing datasets, runs, publication status, and available actions.
Admin dashboard
Atlas DataFlow public content editor for managing the presentation of the Dry Bean dataset study.
Public content editor

Current scope

The first public version of Atlas already covers four problem families:

  • binary classification
  • multiclass classification
  • continuous regression
  • univariate forecasting

These capabilities are currently represented by the Telco Customer Churn, Dry Bean, Concrete Compressive Strength, and Nottingham Monthly Temperatures studies.

Why this project matters

Atlas DataFlow is where my interests in data analysis, machine learning, and software engineering come together. It works both as a study platform and as a portfolio project, allowing me to publish analyses with more structure, usability, and governance than an isolated collection of notebooks.