AI of the Beholder

~ 9 min read

We hear a lot nowadays about the many superpowers of AI, but we don't hear much about its highly refined aesthetic sensibilities. Well, that's what I'm here to change! I trained a convolutional neural net to look through thousands of photos taken from on top of the space needle and learn which ones are junk (as a baseline), and even to highlight the most beautiful of the bunch! But while the concept may be a bit of pet-project-floof, it's not without use case: think first-pass filter on a photographer's raw photos; think auto-curated album from your snap-happy vacation. Read on for more!


Speakeasy, the AI Bartender

~ 15 min read

You know that thing where you're hanging out in a schmancy speakeasy and the bartender asks you what you'd like to have—not in terms of a specific cocktail, or even the base spirit, but in terms of the flavor profile? And then just sets to work grabbing one bottle after another until before you know it you've got a little bit of magic in your mouth and you don't even know how? That. That right there is the epitome of mixology, as far as I'm concerned. That's "the speakeasy experience." That's what I've sought to recreate with this app.



Using Machine Learning to Predict Flight Cancellations

~ 13 min read

Over 99% of scheduled domestic flights take off. That remaining 1% is very expensive, however, by some estimates costing the airline industry nearly a billion dollars every year. Can we use machine learning to predict these cancellations in advance to reduce the associated losses?


Scraping Leafly.com for Data on the Marijuana Industry

~ 5 min read

Leafly is an information aggregator for cannabis. They maintain a profile for most of the dispensaries in the state. This makes them a pretty valuable resource for studies on the cannabis industry. Here's how I scraped Leafly for ratings, reviews, inventory counts, and other metadata on every dispensary in the state of Washington.


Mary Jane, the Model: Using Machine Learning to Predict Marijuana Dispensary Performance

~ 14 min read

In 2012, Washington state passed I-502 and legalized the recreational sale, use, and possession of marijuana. This event has led to an explosion of development in the field that's making waves through our society. Since 2014, approximately 500 state licensed dispensaries have opened throughout the state, with nearly 150 of those here in Seattle. In this project I scour the web for publicly available data that might be predictive of how a cannabis dispensary performs, such as customer reviews, inventory distributions, and local demographics. I then train machine learning models to predict a dispensary's monthly revenue and analyze the resulting models to distill insights about what drives sales in the marijuana market.