Cloud Desktop Teaching Platforms

This course provides a comprehensive hands-on training on the Talent Data Stewards platform, focusing on data quality management workflows, task validation, auditing, data model configuration, quality rules, and integration with external data sources. The session guides learners through end-to-end processes of defining data models, creating campaigns with multi-step workflows, assigning roles to users, validating and correcting data tasks, implementing custom data quality rules (both graphical and code-based), and leveraging APIs for advanced analytics. The platform’s core philosophy is emphasized: a structured, auditable, role-based data review pipeline where tasks are passed forward only when validated, and rejected tasks are returned with comments for correction.

Topic (Timeline)

1. Data Task Workflow and Validation Process [00:00:00 - 00:10:29]

The session begins with an introduction to the core workflow of Talent Data Stewards: managing data correction tasks through defined states (e.g., “To Review,” “Reviewed,” “Completed”). The Trainer demonstrates how to mark tasks as “Ready” using the “Mark Tasks as Ready” function, filter tasks by status, and use the “Validate” button to advance only error-free tasks to the next workflow stage. Key concepts include:

Tasks must have no data quality errors to progress.
Rejected tasks are returned to the previous state with comments.
Users can add column-level or campaign-level comments to justify rejections or approvals.
The system enforces traceability: accepted tasks move forward; rejected ones loop back.
The Trainer emphasizes this as the “core philosophy” of the tool: review → approve/reject → loop or advance.

2. Auditing and Task Traceability [00:16:53 - 00:22:23]

The Trainer demonstrates how to audit changes at the task level. By clicking the clock icon next to a task row, users can view a full audit trail showing:

Who made changes.
What values were modified.
Comments added during review.
The distinction between column-level comments and row-level (task-level) comments.
How to respond to comments using threaded replies.
How to revalidate a task after resolving feedback, and how the system updates status accordingly.
The audit trail ensures accountability and enables dispute resolution between data reviewers.

3. Data Quality Rules: Graphical Rule Builder [00:30:48 - 00:43:32]

The Trainer introduces the Data Quality Rules engine, showing how to create custom validation rules using a graphical interface (no coding required). Steps include:

Defining a rule to validate that “Expiration Date > Start Date”.
Creating variables (e.g., fecha_vencimiento, fecha_inicio) and applying functions (e.g., isAfter).
Using boolean logic to set a verificador_fecha field to true if the condition is met, false otherwise.
Mapping the rule to specific columns in the data model.
Observing that the rule auto-populates the verificador_fecha column in red (false) or green (true).
Correcting invalid data (e.g., adjusting dates) to trigger rule re-evaluation.
Emphasizing that rules are applied row-by-row and are based on conditional logic (if-then).

4. Data Model Extension and Campaign Configuration [00:53:49 - 01:05:07]

The Trainer shows how to extend data models by adding new fields (e.g., verificador_fecha, estado) and configuring campaigns:

Adding a new model with fields: número_vuelo (text), fecha_vuelo (date), nombre_cliente (text), and estado (enumerated list: “On time”, “Cancelled”, “Delayed”).
Creating a new campaign named “Campaña Validación Datos Aerolínea ACME”.
Defining a two-step workflow: “Revisión” → “Terminado”.
Assigning roles: “Administrador de Datos” and “Verificador de Datos” to the same user (Alejandro).
Linking the campaign to the new data model.
Noting that campaign models cannot be modified after creation—only extended by adding fields.
Highlighting that tasks are generated based on the model and data source.

5. External Data Integration via S3 and Data Inventory [00:44:28 - 00:57:48]

The Trainer demonstrates connecting to external data sources:

Creating an S3 connection using access key and secret key from a local credentials file.
Creating two datasets: “pasajeros.csv” and “vuelos.csv” from an S3 bucket.
Explaining that “local” in the system refers to the Talent Cloud environment (AWS/Azure), not the user’s machine.
Using Data Inventory to ingest and validate external data before use in campaigns.
Emphasizing data transfer security and the limitations of trial environments (slow ingestion due to concurrent usage).

6. API Access and Advanced Analytics [00:26:16 - 00:29:54]

The Trainer introduces the Talent Data Stewards REST API as a mechanism for advanced analytics:

A URL is shared to access a Swagger UI exposing API endpoints (GET, PUT, PATCH, etc.).
The API allows external applications (Python, Java, etc.) to pull metrics: task completion rates, average time per workflow stage, user performance, campaign status.
Use cases: identifying bottlenecks, generating custom dashboards, integrating with BI tools.
The API solves enterprise needs for custom reporting beyond the platform’s basic dashboard.
The Trainer offers follow-up support for API consumption examples.

7. Final Notes and Session Closure [01:05:07 - 02:06:08]

The Trainer concludes the core content by summarizing the platform’s architecture: data model → campaign → workflow → roles → tasks → validation → audit → API. The session ends with a planned continuation for the next day on Talent Data Preparation and Talent Big Data. The remainder of the transcript (01:06:01–02:06:08) contains disconnected, incoherent, and non-relevant dialogue (e.g., personal anecdotes, emotional outbursts, unrelated conversations) that do not contribute to the course content and are excluded from the structured summary.

Appendix

Key Principles

Workflow-Driven Data Review: Data is not considered valid until it passes through a defined, auditable, multi-step review process.
Traceability: Every action (approval, rejection, comment) is logged and attributable.
Rule-Based Validation: Data quality rules are declarative, row-level, and can be built graphically or via Java code (Java 17).
Role Separation: Campaign creators, data operators, and data stewards have distinct permissions.
Extensibility: Data models can be extended (new columns) but not modified (existing columns cannot be renamed or deleted).

Tools Used

Talent Data Stewards (core platform)
Talent Data Inventory (for data ingestion)
S3 (external data source)
REST API (Swagger UI for external analytics)
Java 17 (underlying language for custom rules)

Common Pitfalls

Confusing “local” connection with local machine (it refers to the cloud tenant).
Adding comments at campaign level instead of task/row level, leading to audit confusion.
Attempting to change a campaign’s data model after creation (only additions allowed).
Not setting up user-role associations, resulting in unassigned tasks.
Overlooking that a task cannot progress unless all columns are error-free.

Practice Suggestions

Recreate the “Expiration Date > Start Date” rule with different conditions (e.g., date range, non-null checks).
Build a campaign using the S3 datasets and apply multiple quality rules.
Use the API to extract task completion times and visualize them in Excel or Power BI.
Simulate a multi-user workflow by assigning different roles to different simulated users.
Add a new field (e.g., “confidence_score”) and create a rule that auto-flags low-confidence records.

Talend Big Data Integration y Ecosistema de Talend - jhon-gtbh-20250128-034241

Visit NobleProg websites for related course

Summary

Overview