I'm a hands-on data engineer and solution architect working on Snowflake, based in Brisbane. I like taking a real problem from the first business conversation through to a design and a working build.
My recent builds explore one theme: right-sized data governance for teams that rely on subject matter experts rather than a dedicated governance function. Each one is a stage of the same arc.
| Stage | Project | What it does | Built with |
|---|---|---|---|
| Capture | Requirements Accelerator (demo video) | A self-serve interview that turns a vague data request into a plain-language business ask and a technical spec, and says honestly whether it's ready to build | Python, Streamlit in Snowflake, Cortex AI |
| Use | meaning-to-semantic-view (in progress) | Captures what business terms actually mean, in the experts' own words, and turns it into a validated Snowflake semantic view | Semantic views, Cortex Analyst, Snowpark |
| Share | (next) | Governed external sharing with row access policies | Secure sharing, row access policies |
And one for range:
| Project | What it does | Built with |
|---|---|---|
| Listening Lens (demo video) | Eight years of my Spotify history as a Snowflake Native App, including the consumer bind path through references | Native Apps Framework, Streamlit in Snowflake, Cortex AI |
And reusable patterns, packaged as agent skills:
| Skill | What it does | Built with |
|---|---|---|
| Snapshot CDC delta pattern | Turns a curated view into an append-only, action-tagged delta table for reverse ETL, with field-level change history, deletes and retention built in. Worked example verified live | Snowflake Scripting, tasks, Cortex Code skills |
| Row access mapping pattern | Row-level security for one shared table: mapping table plus row access policy, CSV onboarding and offboarding with validation and an audit trail, and a governed copy that survives the source table being rebuilt. Both worked examples verified live | Row access policies, Snowflake Scripting, tasks, Cortex Code skills |
- Problem first. Every project starts from a real problem and documents the decisions behind it, not just the code.
- Claims I can defend. READMEs state what was actually verified, and where. Unverified numbers stay flagged.
- Built with AI, disclosed openly. I use Claude as a drafting and review partner; the problem framing, design calls and scope decisions are mine, and each repo says how it was built.