Hello there! Wouldn't want to waste your time rambling on and on and on....so let's get started!
My name's Gabriel, I'm a junior software developer who studied CS at Princeton (graduated in 2020).
On this page you can find some of my coding projects, hobbies, and other miscellaneous information
that is likely of no use to you. Enjoy!
TigerEats is a web app I developed with some
friends in college. The app helped university
nutrionists manage their students' data and served as an auxiliary
component to the main TigerEats phone application, which focused more
so on providing students with informed meal choices on dining hall food.
The app was built on a React + Flask + MongoDB stack, which all of us contributed to.
I also took ownership over the user progress page – please see our
demo app (I apologize in advance for
Heroku's slow loading time).
LiDAR Perception
During the summer of 2019, I interned at
Hesai, a Shanghai-based LiDAR
manufacturer that supplies perception technology for self-driving vehicles.
As a software engineer on the perception team, I helped build out the objection
detection pipeline for LiDAR-generated point clouds. Sorry, I don't think I'm
allowed to say much more than that!
Chinese Flashcards
I grew fascinated with China after visiting for the first time in 2018
to take a course on contemporary China. When I went back to school in the
fall I decided to formally learn Mandarin. One tool that I found super useful
was an app called Pleco, which allowed me to input words as flashcards for
future study (alongside other cool features like a CHI-ENG dictionary,
character stroke order animation, OCR/handwritten character recognition etc.)
Due to the sheer number of words we learned per unit, manually typing these
in was quite tedious. As a result, I decided to create a python app that
used OCR (and some other custom heuristics) to parse a PDF verison of the textbook
and output the flashcards. You can find the github repo here.
Gulf of Mexico Coral Research
I figured I could use my free time after graduating college to do some social good.
I stumbled on a really cool opportunity to conduct research at Rice University's
Climate, Water and Energy Lab under the direction of Professor Sylvia Dee. The research
project involved identifying factors that were harming corals in the Gulf of Mexico
and using this data to find a way to save this endangered ecosystem. I developed some
Python tools for calculating and visualizing relevant data. You can find the repo
here. A research paper should also be
coming out soon!
Human Skin Segmentation
As part of my computer vision class final project, I worked on a human skin detection CNN for video sequences.
Now, you might be thinking: "that's a pretty niche topic." And you would be correct. I honestly
don't have too much reason for choosing this project other than I thought it would be cool. I trained
this CNN to accept multiple frames as input
and create a refined probability map via a skip layer. This produced state-of-the-art results, although
the detector was far from perfect. You can find the repo, which includes a written
paper, here.
Depth Estimation
I completed an independent work project on the topic of 2D depth estimation.
In short, given some image, is it possible to estimate a relative depth map?
I used a base network to create an initial estimate, then allowed users to
add additional ordinal constraints by clicking on pixels that were incorrectly
estimated. This data would then be used to refine the initial output. Although
the refinement step only improved the estimate when a specific set of criteria
were met, I was pretty satisfied with the results as this was my first
end-to-end machine learning project. You can read more about it on the github
repo here.
Excel Product Management Add-In
The summer after my freshman year of college, I interned with the
anti-money laundering (AML) division of the Bank of Montreal. My team's
role was to manage all the AML projects. Data ingest, calculation,
filtering, updating, migration, presentation etc. were all done manually,
and this was a major pain point for the team, especially since changes
needed to be made often. I coded an Excel add-in that did all
of the above seamlessly. With a click of
a button, new sharepoint data would be added to the master sheet, which
would in turn produce updated calculations on data forecasts.
Moreover, users could easily display views of the filtered data
they wanted to see, which was a common bottleneck during meetings. I even
added an Excel to PPT migration tool that created a Powerpoint presentation
based on the most recent data. All in all, this software enabled the AML group
to present their findings to the board a month earlier than usual.
Computer Music
I took an interesting class on computer music on college, where I got to learn
about sound processing and gained hands-on experience composing various styles
of computer music using a variety of technologies such as ChucK, Max 8,
MobMuPlat and Pd. You can find my projects on
github. Here's a
a celtic inspired sample I made with ChucK.
elecTrip
elecTrip was the capstone project of a tech entrepeneurship class I took my senior year.
I and three other teammates came up with a startup idea that capitalizes on the fast-growing
electric vehicle market. Our paper can be found
here.
Our presentation, which scored among the top two in the entire class
according to students and professional venture capitalists alike (judges: Shana Fisher, Rebecca Kaden), can be found
here.
