Problem
Most food recommendation systems feel opaque and engagement-driven. Users see suggestions, but not the taste logic or tradeoffs behind them.
← Back to projects
Open Demo
2026 live
Flavor Compass
Taste-profile recommendation engine: enter your top dishes and get high-confidence + exploratory meal suggestions with transparent reasoning.
- Astro
- TypeScript
- Recommendation engine
- Explainable scoring
- Product strategy
Solution
Built an explainable recommendation engine that maps favorite dishes into taste vectors, then balances fit vs novelty with controllable exploration settings.
Impact
Showcases product thinking in one interface: user input quality, transparent algorithmic reasoning, controllable risk, and actionable recommendations instead of black-box results.
Flavor Compass is a product-thinking demo disguised as a food tool.
Open the demo: Flavor Compass.
Product principles behind the app
- Explainability first: every recommendation includes “why this fits” and “what’s new”.
- User-controlled exploration: a single novelty slider shifts between safe picks and adventurous picks.
- Actionable output: recommendations are presented as concrete next dishes, not abstract clusters.
Recommendation engine model
- Collect top dishes (user signal)
- Infer a weighted taste vector (heat, richness, acidity, sweetness, etc.)
- Score candidates by similarity and diversity
- Blend exploitation + exploration based on user controls
Why this belongs in the portfolio
This project demonstrates the end-to-end product loop: problem framing, interaction design, algorithm design, and decision UX that communicates tradeoffs instead of hiding them.
Demo Mirror
Live Preview
Mini preview of the actual demo. Use the launch button for full-screen interaction.
