It started with a text from my wife in a grocery store. She didn't know which wine to buy, and I used to live in Sonoma, so apparently I was the person to ask. What I actually wanted was a way to turn someone's palate into a bottle. You describe what you taste, what you're eating, or just "something red and jammy," and Winosaurus searches a cellar of reviewed wines the way a sommelier would, minus the raised eyebrow. The iPhone app does that in your pocket. Scan a label, save the ones you love, and it starts recommending from what you've actually liked. The models behind these have been in development since 2017 when I first built the recommender using a Flask web app and Latent Semantic Analysis. It has gotten way more sophisticated over the years and now involves hand tuned embeddings and graph/network models weeeeeeee!