Machine Learning Engineer at Coframe ↗
A LITTLE WORLD, LEARNING YOU
Meet someone you like. Watch your world become a little more you.
Your kind of curious.
Tap a creature to say hello. Swipe to meet the next one.
A little blue friend, just for you. Enable JavaScript to explore all 48 creatures and watch your preferences shape the world.
A little wonder for Everest & Magnolia.
Every hello teaches this little world what catches your eye. Your next discoveries change immediately.
No assumptions yet. Start with a creature you like.
An online linear preference model learns weights for five curated attributes: element, color, creature type, mood, and material. Each tap makes a positive logistic update. A deliberate hover is a smaller signal. Unselected creatures are not treated as dislikes.
Recommendations balance that score with variety and an exploration slot every fifth arrival. A small weight decay lets your interests change. These are affinity scores, not measured probabilities or claims about your personality.
This is my independent demo, inspired by the adaptive experiences I work on at Coframe—not Coframe’s production model. The artwork is pre-generated; the preference updates happen live in your browser.
Your picks stay in this browser. No account, no analytics, no model request. Reset whenever you want.
The light follows your local clock. Pick a mood, or invite a little weather in.
Optional. City searches go to Open-Meteo. No location permission needed.
Light and weather change the atmosphere, never your taste profile. Motion is optional; swipe and tap always work. Weather by Open-Meteo ↗
Pinch or use + / − to zoom. Drag to explore. Double-tap to zoom or fit.
