Experience
- Own the marketing analytics data models in dbt, consolidating 40+ ad-hoc SQL scripts into a tested, documented layer refreshed hourly for 60 stakeholders.
- Cut the weekly cohort-retention report from a 6-hour manual pull to a 3-minute scheduled run by rebuilding it as incremental dbt models on BigQuery.
- Partner with lifecycle marketing on a shared metrics layer, taking conflicting revenue numbers across teams from a recurring dispute to zero in quarterly reviews.
- Introduced data-quality tests that catch schema drift before it reaches dashboards, cutting broken-dashboard incidents by roughly 70%.
- Built the self-serve returns dashboard in Tableau used daily by a 25-person logistics team, retiring a spreadsheet that cost two analysts a full day each week.
- Ran the A/B analysis for the checkout redesign, reading out to product weekly and confirming a 4.2% lift in completed orders at 95% confidence.
- Automated the weekly returns-rate report in Python, saving an estimated 6 analyst-hours per week.
- Maintained 30+ Looker dashboards for the customer-operations team, turning around ad-hoc SQL requests with a median under one day.
- Segmented 1.2M active customers by support-contact behaviour, informing a staffing model that trimmed average queue time by 18%.
- Analysed booking-funnel drop-off across 12 markets, surfacing a mobile checkout bug that accounted for roughly 8% of abandoned bookings.
- Prototyped a supplier-performance scorecard in SQL and Tableau later adopted by the marketplace team.
- Wrote R simulations for a survival-analysis study, cutting a benchmark runtime from 40 minutes to under 5 with vectorised code.
- Packaged the reproducible-analysis appendix for a submitted paper so every result rebuilt from raw data with a single command.
Projects
An open analysis of a decade of Berlin rent data across all 12 districts, with an interactive Folium map of price changes and a reproducible scraping-to-visualisation pipeline. The repo has 300+ stars and was picked up by a local data-journalism newsletter.
A small app that turns a year of streaming history into a personal year-in-review, built to learn dbt and Tableau end to end. It models raw play events into listening habits by hour, day, and genre.
A DuckDB and dbt project that turns the German census into queryable, documented models, with a Streamlit front end for exploring demographics by postcode. It runs entirely in the browser through WASM, so there is no backend to host.
A weekly demand-forecasting notebook benchmarking Prophet against a gradient-boosted baseline on two years of retail sales, with a plain write-up of where each model wins and why.
Skills
Education
Certifications
Awards
Languages
Volunteering
- Mentor a cohort of women learning Python and SQL, running fortnightly office hours and reviewing their first data projects.
- Have guided 20+ mentees through their first analytics portfolio piece, with several going on to junior data roles.