• 🔧 Analytics Engineer — transforming raw data into reliable, well-structured pipelines and models
• 📊 Passionate about clean data architecture, good documentation, and making data trustworthy
• 🌱 Currently learning: dbt · Microsoft Fabric · Databricks · Python & Data Pipelines · Claude & AI
• 💬 Ask me about Analytics Engineering, SQL, dbt, Power BI and data modelling
• 📝 Writing about data & tech on Shift with Jo

Transform & Model

Platforms

Tools

Capstone project — Data Engineering track, Data Girls Bootcamp 2026.

An end-to-end ETL pipeline that extracts a credit score dataset from Kaggle, cleans and transforms it with pandas, and loads the result into a Databricks Unity Catalog Volume via the Files API.

• Transform: handles corrupted/sentinel values, PII pseudonymization (hashing + reversible ID mapping), type casting, range checks, and outlier/missing-value treatment — with the target column always kept untouched
• Orchestration: Apache Airflow running in Docker, backed by a Postgres metadata database so DAG run history and the admin user survive container rebuilds
• Scheduling: a daily DAG with automatic retries and a failure callback that logs which task/run failed and where to find the logs
• Documentation: every data-quality and architectural decision is explained and justified in the README

Python pandas Airflow Docker Postgres Databricks

Capstone project — dbt/Analytics Engineering course. A dbt project built on Snowflake from scratch: a medallion-architecture pipeline (raw → staging/bronze → silver) over real public aviation data (~72K airports, ~44K runways, and user comments from OurAirports, joined on the airport_ident ICAO code).

• Incremental models: built to process new/changed data efficiently rather than full-rebuild every run
• Snapshots: slowly changing dimensions (SCD Type 2) tracked via dbt's built-in snapshot feature
• Layered testing: generic, domain-specific, cross-model, and singular tests, extended with dbt-expectations, plus test-failure persistence for debugging
• Documentation: doc() blocks, full model/column descriptions, and an interconnection overview tying the whole DAG together
• Environment: managed with uv for reproducible Python/dbt tooling

dbt-core Snowflake dbt-expectations SQL uv

Three independent, self-contained data projects, cleaning, static reporting, and a live BI dashboard; each solving the same class of e-commerce/CRM problem with a different final delivery format.

A portfolio repo where every project lives in its own folder with no shared code, its own dataset, and its own README, so eacendently.

• Sales data cleaning: Streamlit app + CLI that diagnoses and fixes a CRM sales export, leaked status fields, flagged/contet capitalization and accentuation, mixed date formats, revenue as BR/US-formatted text, with genuinely missing or unparseable values left untouched and surfaced for manual review, never invented.
• Static analytics report: a CSV → JSON → HTML pipeline that computes e-commerce KPIs (revenue trends, category/brand bregaps, geographic distribution) into small intermediate JSON files and renders a single static report, keeping the ~3,000-row sales table out of memory/context all at once.
• Live analytics dashboard: a multi-page Streamlit dashboard over the same dataset served from a Supabase (Postgres) database read-only via the anon key — Sales, Price Positioning, and Customers pages each backed by a pure, unit-tested KPI module, with every formula and data-quality caveat documented.

Python pandas Streamlit Supabase pytest