Practical prototypes connecting data ownership, AI risk controls and traceable workflows.
These projects explore how governance requirements translate into inspectable rules, evidence and applications.
A runnable Python and SQL reference pipeline that turns synthetic claims-payment data into a governed daily data product. Versioned contracts, quality gates, quarantine and replay-safe loading connect source records to traceable metrics.
Verified: 15 behavioral tests, sample ingestion and evidence export passed in GitHub Actions on Python 3.11, 3.12 and 3.13. Implemented with SQLite; PostgreSQL, dbt and cloud deployment remain planned extensions.
Walkthrough · Architecture · Validation
A synthetic reference implementation covering AI inventories, configurable risk classification, lifecycle gates, monitoring and oversight. Includes Python control logic, YAML rules, tests and a Streamlit demonstration.
A fictional operating model connecting data domains, ownership, business definitions, quality rules and governance issues. Includes structured registers, validation tests and a Streamlit portal.
A document-retrieval prototype using synthetic ESG material, TF-IDF search and source references. Supports an extractive demonstration mode and optional LLM-generated drafts, with an insufficient-evidence threshold and human review.
- Explicit ownership and reviewable governance decisions
- Versioned data contracts, quality gates and replay-safe pipelines
- Configurable rules and validation controls
- Source traceability and evidence for human review
- Python, SQL, YAML, Streamlit, pytest and GitHub Actions across the portfolio
These are demonstration projects. Scope and limitations are documented in each repository.