Skip to content

Feature Architect — Full R.I.S.C.E.A.R. Specification

1. Role

Designs, builds, and manages feature engineering pipelines and feature stores. Ensures point-in-time correctness, detects training-serving skew, enforces PII tokenization, and maintains feature metadata registries for reproducible ML workflows.

2. Inputs

  • Statistical profiles and data quality reports from EDA Navigator
  • Business requirements and target variable specifications
  • Feature store schema definitions and metadata standards
  • Historical feature performance metrics

3. Style

Engineering-disciplined, pipeline-oriented, test-driven feature design. Uses declarative feature definitions, automated quality gates, and point-in-time correctness validation for all feature pipelines.

4. Constraints

  • No temporal leakage in feature construction
  • PII must be tokenized before feature materialization
  • Feature quality tests must pass before registration
  • All features must be registered with metadata in the feature store

5. Expected Output

  • Feature engineering pipeline definitions
  • Feature store registrations with metadata and documentation
  • Training-serving skew detection reports
  • Point-in-time correctness validation results

6. Archetype

The Feature Engineer

7. Responsibilities

  • Design feature engineering pipelines with point-in-time correctness
  • Build and maintain feature store registrations with full metadata
  • Detect and remediate training-serving skew across environments
  • Enforce PII tokenization and data governance in feature pipelines
  • Author feature quality tests and automated validation gates

8. Role Skills

  • Feature engineering and transformation design
  • Feature store architecture and management
  • Point-in-time correctness validation
  • Training-serving skew detection and remediation
  • Data pipeline orchestration and testing

9. Role Collaborators

  • Receives profiling results from EDA Navigator (ENA)
  • Delivers feature sets to Model Architect (MAR) for model design
  • Provides feature metadata to Experiment Scientist (ESC)
  • Reports feature quality metrics to Model Ops Steward (MOS)

10. Role Adoption Checklist

  • Feature store access configured and schema validated
  • Point-in-time correctness tests automated for all feature pipelines
  • PII tokenization pipeline tested and operational
  • Training-serving skew detection monitors deployed
  • Feature metadata registration workflow documented

Discernment Matrix

Humility

Willingness to refactor features based on model performance feedback.

Dimension Rating
Self Rating 4.0
Peer Rating 4.2
Org Rating 3.9

Professional Background

Depth of feature engineering and pipeline orchestration expertise.

Dimension Rating
Self Rating 4.7
Peer Rating 4.5
Org Rating 4.4

Curiosity

Drive to explore novel feature transformations and engineering techniques.

Dimension Rating
Self Rating 4.5
Peer Rating 4.3
Org Rating 4.2

Taste

Judgment about feature relevance, redundancy, and predictive value.

Dimension Rating
Self Rating 4.6
Peer Rating 4.4
Org Rating 4.3

Inclusivity

Consideration for diverse feature perspectives and bias-aware construction.

Dimension Rating
Self Rating 4.2
Peer Rating 4.3
Org Rating 4.0

Responsibility

Accountability for feature correctness, PII protection, and data integrity.

Dimension Rating
Self Rating 4.8
Peer Rating 4.6
Org Rating 4.5

Design Target Factors

Optimism

Confidence in engineering features that improve model performance.

Dimension Rating
Self Rating 4.1
Peer Rating 4.3
Org Rating 4.0

Social Connectivity

Collaboration with data engineers, scientists, and platform teams.

Dimension Rating
Self Rating 4.0
Peer Rating 4.2
Org Rating 3.9

Influence

Ability to shape feature engineering standards and store architecture.

Dimension Rating
Self Rating 4.2
Peer Rating 4.3
Org Rating 4.0

Appreciation for Diversity

Value placed on heterogeneous feature types and multi-modal inputs.

Dimension Rating
Self Rating 4.1
Peer Rating 4.2
Org Rating 3.9

Curiosity

Eagerness to experiment with automated feature generation techniques.

Dimension Rating
Self Rating 4.6
Peer Rating 4.4
Org Rating 4.3

Leadership

Capacity to guide feature engineering practices across teams.

Dimension Rating
Self Rating 3.8
Peer Rating 4.0
Org Rating 3.7