I code custom software solutions for various companies. My favorite work involves scraping massive amounts of potentially pertinent reddit data (Selenium crawling -> Python requests -> Reddit JSON endpoints) then using batch jobs to get LLMs to filter and categorize the data into consumer insights for various products and services.

Developed embedded Google Payments software using a proprietary Java SSR framework, integrated into Google Play and YouTube platforms (Jul 2021 - Feb 2023).

Engineered features for next-generation Google Smart Displays (Dart/Flutter) and Home Automation Routines in the Google Home App (JavaScript/TypeScript) (Oct 2019 - Dec 2020).

As one of the early engineers in this SF-born startup, I focused on developing new features such as a cart, new UI, and updating the buying experience for a React Native app focused on social buying and selling of recommended products.

At 16 I dropped out of high-school and moved to CA to start my own company. Both attempts (Cohereum & Levana) were focused on coding an Auto-ML platform to make building custom machine learning solutions easier for non-tech businesses. We raised a few million USD and developed an end-to-end platform for building neural nets, but ultimately failed. Nonetheless, the experience taught me a great deal about perseverance, company structure, law, technology, and machine learning.

A ML approach to measuring music similarity I built for a friend to increase click-through rates and decrease advertising cost for their music. Theia crawls YouTube Music downloading songs to compare to a baseline spitting out the top-N closest songs. Theia decides "similarity" by using the cosine distance between the second-to-last layer of a genre classification Convolutional Neural Net that takes in the music spectrograms as visual input data.

My longest running and favorite project is my marriage to my wife, Xinyu, who've I've been with for over 5 years at the time of writing this. Our love is memorialized in a memory book website I coded. Check it out to see our cute memories!:)

My personal website - the one you're currently viewing. The thumbnail is the last version. I host it on Netlify, and have it open-source on Github. I have had versions built using all sorts of JS frameworks and libs (Vue, React, JQuery), but ultimately I think plain HTML/CSS/JS is just fine given the requirements.

A little - but fun - game I made entirely in VueJS. Fight monsters, buy new spells, find items, travel, build, a whole lot more in this little SPA game! I originally made it for task management but it became a game.

BEFORE GPTs EXISTED: PyTinder was a fun side-project that aimed to automate the entire Tinder lifecycle: login, swiping right, initiating and responding to chats, and getting the date being setup. PyTinder uses Selenium to automate interactions with the Tinder web interface and my own trained Seq2Seq LSTM (Before GPTs existed!) to chat with matched users.