Willie
Persoonlijke wijnaanbevelingen op BigQuery + dbt
(Engelstalig origineel — Nederlandse vertaling volgt.)
What it does
Willie is a recommendation engine for the Dutch wine retailer wijnboodschap.nl. Given a customer's order history, it surfaces wines they will probably like — but with explanations the customer can read out loud:
- Geographic match — same region or sub-region as a wine they bought before.
- Vintage year overlap — birth years, anniversaries, milestone years.
- Name overlap — playful matches based on shared words or themes.
Each recommendation comes with a one-sentence reason — no opaque "because the algorithm said so."
Architecture
- BigQuery + dbt — the wine catalogue and order history live in BigQuery. dbt transforms them into the feature tables Willie reads.
- Python / Flask backend — small API that serves recommendations.
- Static front-end — embeds into the wijnboodschap.nl site.
Status
In dev. The backbone (BigQuery + dbt) is up; the front-end embed is the active build target.