Building modern risk infrastructure

Context

The project began with risk data and processes distributed across Excel files and a legacy SQL database—a practical starting point that made consistent history, freshness, lineage, and reuse increasingly important.

How I approach it

Within a wider team project, I designed and implemented core elements of the migration: Snowflake RAW and historical layers, governed views, and dispatcher-led orchestration. I also built automated data-quality and freshness checks and a standardised component layer for risk and performance dashboards. Semantic structures extend the same governed definitions and data relationships to the AI risk agent.

• Governed Snowflake RAW and historical layers with curated views
• Dispatcher-led orchestration with automated quality and freshness controls
• Reusable dashboards and semantic structures shared with the AI risk agent

Why it matters

Once definitions, calculations, lineage, and controls live in a reusable analytical layer, they do not need to be rebuilt separately for dashboards, reports, or AI. People and models can work from the same trusted foundation: deterministic interfaces support monitoring and analysis, while the agent uses that context for investigation and synthesis. That makes AI part of the analytical system rather than an addition to disconnected data.