Projects

Open-source (MIT) desktop app that treats a Power BI / Fabric semantic model as an API. Author a dashboard once — in a visual editor or from a plain-English prompt — compile it to a portable App Spec (JSON), then render that same spec on four surfaces: the desktop app, a self-hostable web app, a custom .pbiviz inside Power BI, and Fabric. No data copy and no backend: DAX runs against the live model through executeQueries, so measures and row-level security behave exactly as they do in Power BI. 30+ ECharts visual types with cross-filtering and drill-down; bring your own key for Anthropic / OpenAI / Gemini / Groq. Public beta at v1.0.0-beta.2.

Streamlit app that compares a source dataset (raw / upstream / app extract) against a target dataset (curated / warehouse / report extract) — file-level CSV / XLSX inputs up to 200 MB, with a Validator pass and a Mismatch Explorer for row-level drill-down. Built for the moment analysts and BI engineers need a fast, no-storage sanity check during ETL handoffs or report-vs-warehouse reconciliations. Public live demo is gated by a private access code — ping me for one.

Public Microsoft Fabric reference implementation: a fictional mid-market apparel retailer with a full medallion-architecture pipeline (Bronze → Silver → Gold, 14 fact and dimension tables). 306 / 306 generator tests passing, production Fabric infrastructure across three workspaces and two service principals, CI via GitHub Actions. Addresses a community need: realistic, license-free retail datasets with intentional data-quality issues for teaching and tool benchmarking.

Open-source educational repo that teaches SQL through narrative storytelling rather than abstract schemas. Progressive curriculum where each chapter advances a story while introducing joins, window functions, CTEs, and performance considerations in realistic business scenarios.