I build agentic systems|

Governance Standards & Frameworks

I define what governed means on both platforms and write it down: the governance framework documentation, separate user and developer guides, the criteria for monitoring and dashboards, and a cross-platform governance strategy. These are standards with consequences, not guidelines — dashboards that fail the criteria lose full platform support.

Compliance Campaigns & Enforcement

The recurring campaigns that hold the estate to those standards: metadata compliance re-checked at the start of each month, results published centrally, and dashboards that have drifted out of compliance identified and followed up. When AI-generated compliance scores were introduced, I verified them by hand against randomly sampled datasets before publishing anything. Taking model output on trust is the fastest way to lose a governance programme's credibility.

Data Quality

Ran data quality campaigns, curated and cleaned critical datasets, traced root causes when columns disappeared from downstream datasets, and set up automated detection of schema changes and deprecated structures — catching issues before they reach consumers.

Data Analysis

Delivered platform and dataset usage analyses, incident and request tracking, cross-platform comparisons, and ontology objects that give engineering teams a reliable picture of their data landscape.

Reporting & Visualization

Built the monitoring dashboards leadership and teams rely on — platform adoption, operational status, incidents, and governance KPIs — turning raw platform data into decisions.

Front-end Development

Developed landing and welcome pages and interactive components (HTML, CSS, JavaScript) so that governance tooling is something colleagues actually want to open.

Platform Operations & Advisory

Taking over operational incident and service request management for the platform, which means challenging requests for redundant structures rather than implementing them as received. Increasingly consulted by engineering teams on metadata questions — classification entered into the wrong fields, gaps that block verification, retrieval impeded by mixed coding schemes — where the ask is judgement rather than build.

Training, Adoption & Continuity

Knowledge transition, onboarding and shadowing, documentation split deliberately into user and developer audiences, and the portfolio KPI reporting that makes the team's work visible upward. The point of all of it is that the governance function outlives any single person — including me.

How a project flows

The team — 11 specialists

How the team works — phases & checkpoints

Every agent works in discrete phases and waits for explicit approval before the next station begins. Nothing is spent on work that has not been confirmed.

Checkpoint behaviour

After completing its deliverable, every agent summarises what was produced, states which station comes next, asks for explicit approval — and stops. The solutions architect goes further: it always presents a two-option brief with an honest trade-off per option before any full proposal is written. Silence or ambiguity is never treated as approval.