Shirley Su

About Me

Hi, my name is Shirley Su. I am a junior at the University of Southern California (USC), pursing my Bachelors in Computer Science.

My passion is to bring proactive change in my community: whether it is creating technology to develop innovative solutions, or volunteering to help out local organizations, I always want to bring positive change. I am especially interested in learning Machine Learning, and its potential to solve several key issues such as climate change.

At USC, I am a developer and VP of Community Outreach at Code the Change (CTC), an organization that drives social change by building applications for LA non-profits, such as ImagineLA and Ron Finley Project.

In my free time, I enjoy playing volleyball, badminton, and swimming. I also enjoy skateboarding around campus. With friends, I enjoy playing card and board games, my favorites include ERS and Catan!

Experience

Aug 2022 - Present

Software Engineer Intern @ TikTok

I am currently a software engineer intern on the Brand Ads team for TikTok. I developed an AMP Boost Platform which would adjust the reach and freqeucy of an Ad. My developement allowed for gradual increases and decreases to the reach and frequency values, and it made it easier to intervene if an Ad's value were not on target. This project was developed using GoLang. My second project was developing a flowmap that mapped the country to the day of the week to a json of Ads that had a refresh type of open. This allows Ads to be better categorized into various refresh types such as open, batch, etc. This project was developed using SQL, Python, and GoLang. My third project was creating an SQL query that backflowed the necessary data to add a new country to a flowmap.

May 2022 - August 2022

Software Engineer Intern @ Microsoft

I worked as a software engineer intern for the Commerce and Ecosystems, Microsoft team. I analyzed MACC Analytics databases to improve accuracy commitments and consumption streams, which lead to a 91.0% to 92.9% MACC accuracy increase. I was able to pinpoint variances among Rating Overages, Commerce Admin, and Macc Services streams to categorize and fix internal issues from incorrect reporting. The languages I used in the project were Kusto Query and Azure Data Explorer.

January 2022 - May 2022

Course Producer @ USC Viterbi

I served as a Course Producer for USC'S CSCI103: Introduction to Computer Science C++ class. I hosted biweekly office hours help answer conceptual questions and debug programming projects and labs. I also led weekly lab section to teach programming fundamentals and grade assignments for 50+ students

June 2020 - August 2020

Software Engineer Intern @ Microsoft

This summer, I served as an Onnx Open Source Contributor. Working with my previous mentors, I deployed EfficientNet-Lite4, an image classification model, as well as SSD-MobilenetV1, an object detection model, to the Onnx Model Zoo. This process involved converting its Tensorflow framework to Onnx format, running validation and inferencing, and writing the preprocessing and postprocessing code for the models. In addition, I also contributed a Model Zoo API, which allows users to download pretrained models from a script and obtains the model's metadata. I created a function that allows the user to download a specificed model. Moreover, I also utilized regular expression and model folder traversal to create the necessary dictionaries to map the model's file name to its download url. ONNX Github

June 2019 - August 2020

Software Engineer Intern @ Microsoft

I served as a Microsoft AI Platform Intern. I worked with the Onnx team to train the Super Resolution model, and then convert it from its Pytorch framework into Onnx format. I used a virtual machine during the training process, which allowed me to gain hands-on experience with working with virtual environments and operating systems. I also helped reorganize the Onnx Model Zoo Github to improve readability and usability. For example, I organized the model folders based on its function, such as classification, object detection and segmentation, machine translation, etc. Moreover, I also improved the descriptions for the model by giving brief details about its accuracy, architecture, and functionalities.

Projects

Ron Finley

On the Code the Change team, we built a Customer Relationship Management (CRM) system for the Ron Finley Project team. Ron Finley Project is an organization dedicated to teaching communities in food deserts how to garden. They constantly need volunteers to help maintain the garden, and our project helps manage this volunteer system. The volunteer manager is able to utilize our CRM to improve communication to better onboard new volunteers, coordinate with old volunteers, and plan future events.

ImagineLA

On the Code the Change team, we developed a full stack web application designed to calculate and manage user's government benefits based on their current income and living conditions. We worked with the ImagineLA nonprofit team to analyze government benefits to develop the backend API. This allows users to be aware of all government benefits they are eligible for and help plan their net income accordingly. Demo

Diagnose Me

Diagnose Me is a web application that enables individuals to self-diagnosis diseases with confidence levels over 400 diseases by allowing them to filter and select from hundreds of common symptoms ranging from the common sicknesses such as the Flu, to the more complex issues such as mental disorders. This project was built with Angular8, TypeScript as the Frontend, and Python (Flask), MongoDB and Google/Symptoms API as the Backend. At the Pearl Hacks 2020, this project won Best Hack to Simplify an Everyday Process. DevPost

Resume

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';