Starbucks
drink recommender

Starbucks
drink recommender

KNN + Content-Based + OpenAI-Augmented Features




I always had a hard time deciding what to order at a coffee shop, so I built this beverage recommender to make my life, and hopefully yours, easier. Give it a try and see what recommendation you get!

How I built the recommendation engine :

  • Collaborative filtering (KNN on 10k+ real customer ratings)

  • Realtime user input for Content-based filtering (caffeine, calories, milk, sweetness)

  • OpenAI-generated taste vectors (GPT-4 distilled 5 latent profiles directly from drink names, they are ruity, refreshing, creamy, earthy/spiced, and bold )



demo

View in GitHub

© 2026 by Nicolie Ng | Data Analyst

© 2026 by Nicolie Ng | Data Analyst

© 2026 by Nicolie Ng | Data Analyst

Create a free website with Framer, the website builder loved by startups, designers and agencies.