Shared method

Value first, then method, then limits.

Every workflow uses the same discipline, not the same model: sources, controls and metrics change with the problem.

Narrative example

How public data becomes a publication.

  1. 1. The workflow acquires the value and its reference date.
  2. 2. It checks missing values, freshness and consistency.
  3. 3. It computes indicators or applies a model.
  4. 4. It compares the output with relevant gates and baselines.
  5. 5. It generates the card, alt text, provenance and duplicate hash.
  6. 6. It publishes or records an explicit skip/block.

From plain language to technical language

The same stages, with two sets of names.

Plain languageTechnical term
The system starts the workTrigger
It retrieves informationAcquisition
It checks that it can be usedValidation / quality gate
It performs calculationsCompute
It decides whether to proceedDecision
It stores results and informationPersistence / provenance

Progressive path

From visible output to implementation.

Each level adds detail without forcing the reader to restart: enter at the most useful point and continue progressively.

Before method

Need a map of AI, data and algorithms?

The intermediate level explains data types, regression, classification, clustering, neural networks and automation.

General diagram

1Public sources
3Cleaning
4Quality checks
5Rules or models
7Output
8Supervision
Modular workflow illustration: sources, controls, model, output and feedback.
The flow is linear in the narrative but cyclical in maintenance: each outcome feeds controls and improvements.

Choose the depth

One method, five reading levels.

The path starts with concepts and examples, moves through operational logic and software architecture, and reaches validation, frameworks and failure modes.

Level 1 — For everyone
  • Where data comes from
  • Why controls matter
  • What a model produces
  • Why supervision remains necessary
Level 2 — Operational
  • Ingestion and normalisation
  • Quality gates and scheduler
  • Error handling
  • Versioning and publication
Level 3 — Technical
  • Feature engineering
  • Temporal splits and baselines
  • Bootstrap and calibration
  • Metrics and drift

RPA in practice

Automation is not only execution: it is exception management.

01

Trigger

Cron, event, manual dispatch or a condition on a source.

02

Acquisition

APIs, feeds, files or pages with retries, timeouts and source dates.

03

Decision

Rules, models, quality gates and supervision determine publish, skip or block.

04

Trace

Logs, state, provenance, hashes and history make the process auditable.

Next step

From workflow logic to software architecture.

The operational level connects triggers, adapters, data contracts, state, tests, CI/CD and publishing.

Open operational architecture