Worked under the Laboratory of Emerging Intelligence on two different projects

AI Tutor Demo:

Developed module to expand the number of tokens for uploaded text files

Improved lecture search feature by 2.3x by enhancing video lecture parsing

Improved image parsing enabling users to upload images instead of text in the demo

AI Tutor Production Software:

Architected and developed production-level software for AI Tutor project.

Utilized Javascript (Svelte) and CSS for the frontend, and Python for the backend.

Led a team of 3 engineers in the development and deployment of the software.

Currently over 400 students get weekly assignments on the AI Tutor software

Designed a predictive model to analyze time series data relating to water well contamination and predict when water levels will reach safe contaminant levels.

As a software engineer intern at MathWorks, I focused on prototyping a new tool requested by clients.

The tool developed would allow engineers and clients to convert Stateflow charts to more traditional flow charts. The tool was created using MATLAB as the intermediate for extracting data from the Stateflow chart and passing it to the Javascript-based web flowchart display. Furthermore, much of the complex algorithmic code was prototyped and tested in Python before being implemented in Javascript. I also implemented bidirectional synchronization to allow any edits made to the Stateflow chart to be rendered in real time on the flowchart, and vice versa.

As a software engineering intern, I focused on the development and enhancement of two internal tools:

Firstly, I developed a tool from scratch that enabled DevOps engineers to track all live and past builds of different modules within the company’s products. This tool was created using a Python-based web server which made API calls to get build information and displayed the relevant data on the portal, and a MongoDB database that stored the build information until it was retrieved to be displayed. The software also implemented a reverse API call which allowed the system-build tool (where data was coming from) to notify the server if a new build was created or a live build was running.

Secondly, I enhanced an internal cost-tracking tool. To achieve this, I used Chart.js as the chart display library, and AWS lambda API calls to access the relevant cost data. The end result was an interactive tool that would show the relative breakdown of factors as they contributed to the cost metrics the company tracked.

As a machine learning intern, I successfully created software using OpenCV (python optimized computer vision library) and Tensorflow (machine learning model training and deployment software) to analyze what letter a child is writing in the air and their movements and facial features.

My software came equipped with active learning, which enabled educators to continue improving the model by inputting images of letters and labeling what letter the image corresponded to.

The software's analysis of facial features and letter writing will give educators the necessary information to be able to recognize symptoms of Tourette's syndrome and down syndrome.