• home
  • bio
  • work
  • blog
  • contact

    • Work.
    Projects.
    Experience.

    Currently a Product Manager @Microsoft. This is list of my notable experiences. For more, follow me below.

    © 2020 Jason Yi
    .001 | Present
    Summary.
    Product Manager.
    @Microsoft
    Currently, within the Azure organization.
    ×
    .002 |
    Summary.
    Software Engineer.
    @Walmart Labs
    Background:  One of Walmart Labs’ primary goals is to improve customer satisfaction. Currently, the wait times are bottle-necked by the customer's actions including pressing the wrong button, not checking in, misplacing their phones, etc. This is where The Walmart Labs Vision team comes in. In order to provide customers a better user experience and obtain more customer analytics, the team was tasked with creating a computer vision proof of concept to detect arrivals at Walmart parking lots and develop a set of criteria for placing cameras in the appropriate locations to collect suitable footage for this solution. Results:  I helped develop the backend computer vision pipeline using neural network models to improve and streamline Walmart’s grocery pickup service. The team primarily used the following pre-trainned models: YOLO, Mask R-CNN, Faster R-CNN. While tweaking the models, we also implemented the Hungarian algorithm to improve the model's efficiency when matching cars with empty parking lots. In order to visualize our results, I programmed a script to visualize the data and access the tradeoffs between the models in the following areas: speed, accuracy, and camera placement features. The Faster R-CNN performed the best with 88% accuracy.
    ×
    .003 |
    Summary.
    Summer Analyst.
    @Bank of America Merrill Lynch
    Background:  As part of Bank of America's summer rotational program, I was placed within the Global Transaction Services organization. Throughout the rotation, I was a part of the following teams: CashPro (BofA's online corporate banking system), Emerging Payments, and Deposit Initiative. Results:  On the CashPro team, I was a Product Manager. I lead a small team of engineers to create a Proof of Concept for a newly designed CashPro system using Blockchain Techology. The new, core feature was a multi-currency account property that allows corporate clients to store their cash flow in different currencies. On the Emerging Payments team, I developed a strategy to partner with international digital wallet companies and incorporate them into our payment rails. On the Deposit Initiatives team, I created data visualization charts to help move the deposits of large financial clients from Deutsche Bank to BAML.
    ×
    .004 |
    Summary.
    Frontend Developer.
    @Market Modeling Project
    Background:  Four friends and I, worked on this project in class. The idea was to help financial analysts find an accuracte prediction of future stock prices, exchange rates, etc. And so, we were hoping to use machine learning and display its' results in a UX friendly manner. Results:  During our research, we mainly compared the results of two recurrent neural network models: Long Short Term Memory (LSTM) and the Echo State Network (ESN). We also focused on building a user friendly website to practice our front end skills. Some frameworks and tools we learned were AJAX, to load data asynchronously and Chartist.js, to help create our pretty graphs. Go check out our Heroku hosted webapp!
    ×
    .005 |
    Summary.
    Designer + Developer.
    @Photo Ally Project
    Idea:  My friends and I decided to join a CalTech Hackathon in 2019. The idea was to develop an IOS app that would recognize pictures of notes, blackboards, and whiteboards and categorize the user's photos into separate albums. This image recognition tool was built using a convolutional neural network model. Go check out the codebase!
    ×
    .006 |
    Summary.
    Researcher + Software Engineer
    @Transcompiler Project
    Background:  One of our CS Profesors told us that American Express was looking for a way to translate SAS to Python language in order to migrate their system. In an attempt to learn more about programming languages, my friends and I decided to work on the project. Results:  After some research, we found a third party parsing tool that generates parse trees from code, called Antlr. And so, we decided to translate a simple 'for' loop from SAS to Python. Once we created the grammars for Antlr, we were able to follow a simple work pattern: SAS -> Parse Tree -> S-Expression -> Python.
    ×