My 7-year experience in AI has made me an end-to-end generalist as I collaboratively designed and built ML systems: from data engineering, ML modelling, backend API development, to DevOps. I started out specializing in Natural Language Processing (NLP) for 5 years before shifting my focus to Geospatial Analytics in the past 2 years. At the end of the day, I love working with a team to build systems that positively impact our end-users.
Almost all of my projects in corporate are (understandably) confidential, so I created this website to share side-projects that are fit for public consumption. I hope you enjoy looking through them as much as I enjoyed creating them!
Click on the project cards for demos, or click on the GitHub links for the code.
![Movie Recommender](/project_imgs/movie_recommender.png)
Movie RecommenderNLP
Find movies to watch by describing plot themes like “scifi stuck in space”, “2 dudes goofing around comedy” and “time travel romance”. This is a full-stack Django app with React embedded within to provide SPA functionality for the main page.
![GitHub](/youtube_icon.png)
![GitHub](/github_icon.png)
Bible SearchNLP
What does the Bible say about ...? This is a semantic search tool to surface relevant verses and chapters in the Bible for a topic of interest using embeddings and Question-Answering models. Yes, this was made before ChatGPT exploded; I worked on it very sparingly on-and-off since 2020. I'm hoping to improve this with ChatGPT's API soon. But if anything, this should run faster and displays whole chapters for better context.
![GitHub](/github_icon.png)
UNICEF AI4D Relative Wealth Mapping for SEAGeo
Relative wealth estimates for 9 Southeast Asian countries. This is similar to Meta's Realtive Wealth Index estimates, but utilizing only open datasets and low-cost Random Forest models. We made this as a baseline work for SEA with an emphasis on replicability to spur further research in the region and beyond.
![GitHub](/github_icon.png)
GeoWranglerGeo
Python library to cut down dev time for geo-data processing and feature engineering so Data Scientists can focus on analysis or ML modelling. We made this to support common geo-data wrangling steps in our client projects, including our SEA Poverty Mapping work with UNICEF. I acted as Product Manager for this project.
![GitHub](/github_icon.png)
![Bible Study Material App](/project_imgs/dg_material.png)
Bible Study Material AppGeneral
React app to digitally serve UECPasay's Discipleship Group material.
![GitHub](/github_icon.png)
![Charge [2013]](/project_imgs/charge.png)
Charge [2013]Game Dev
Published Windows Phone game with 12k+ downloads. Made for a Microsoft hackathon in 2013. This is a video-game re-imagining of the popular Filipino kid game called Charge (played in real life with hand actions). The core concept is to "Charge" up, use your charges to "Wave" and attack your opponent, which you can "Block". Getting hit by a "Wave" from your opponent while you're "Charging" means game over for you. This video-game takes that core concept and turns it into a semi-rhythm game. Much like the rock-paper-scissors chant, you have to tap and choose your action in time with the beat to defeat the AI! Unfortunately, the Windows Phone store has shut down and so the game is not available anymore.
![GitHub](/youtube_icon.png)
![Less than Three (<3) [2013]](/project_imgs/less_than_three.png)
Less than Three (<3) [2013]Game Dev
Javascript game made for a 2013 Valentine's Day Hackathon (winning entry) during our sophomore year as BS Computer Science students. Your objective is to find the one for you given clues to what they're wearing. Once you find them, you have to answer correctly to pass the 'compatibility' test. Every character is dynamically generated for each new run of the game. Warning: The game is unpolished as this was made in ~24 hours and was our first hackathon ever during our 3rd year as BS Computer Science students. No game engine was used; this was pure HTML/CSS/JS!
![GitHub](/youtube_icon.png)
![GitHub](/github_icon.png)