For additional projects, view my GitHub.
Developed a startup from the ideation and customer interviews phase to a launch-viable MVP with a team of 4 as part of my capstone project at Georgia Tech. This startup, 'North* Documentation', focuses on automatic generation of high-quality end-user documentation for web applications using AI. Several technologies were used in this project, including:
The GitHub Application allows users to easily install our documentation generator on any repository. Once installed, users can trigger the documentation generation process with a single click. The application then uses Playwright to scrape and interact with the web application, gathering necessary information for documentation generation.
Connecting GitHub Repository and Starting Documentation Generation
When a repository gets submitted for documentation, our system clones the repository into a temporary folder and uses Docker to host the web application as an isolated container. From there, our AI pipeline iterates through each page of the web application using the following steps. This process repeats until all pages are processed:
Once all pages are processed, the AI compiles the screenshots and instructions into a well-structured markdown format, which can be downloaded and converted to HTML or PDF for distribution. Previewing the documentation is also possible on our system, which uses Nextra to render the markdown and display it to the end-user as seen below:
Previewing Generated Documentation
Developed a prototype for several different types of AI integration using OpenAI's Generative AI and a RAG system for dynamic prompt creation. The different types of AI Integration include feeding a paragraph into OpenAI's API for clarification, feeding user code into OpenAI's API for debugging, and using a RAG system for dynamic context injection. Steps to achieve this included:
The first prototype focused on AI-powered paragraph clarification. It allowed users to click a paragraph and receive a more detailed explanation utilizing previous relevant lessons as context.
AI Paragraph Clarification
The second prototype focused on AI-powered code debugging. It allowed users to click a button in the inline code editor to receive suggestions for fixing errors. The AI utilized the context of the code to determine language and returned suggestions in formatted HTML for easy integration.
AI Code Check
The third prototype focused on AI-powered pseudocode generation. It allowed users to click a button to receive a structured outline of how the current activity could be completed.
AI Pseudo Code Generation
Developed a reporting engine to provide teachers and students insights into their learning progress and areas for improvement. This project involved full-stack development, including complex SQL statements, data visualization, and user interface design. Successfully led this project from start to finish, including concept development, front and back-end implementation, and user testing.
Technologies used for this project include:
An example teacher dashboard may look like this:
Teacher Report Dashboard
The reporting engine also includes a specialized student dashboard, giving more personalized data insights to each individual user. An example student dashboard may look like this:
Student Report Dashboard
Worked with a team of 7 classmates to develop an android app using the Spotify API. This android app allowed users to visualize their music listening habits and compare them with their friends.
This app used several libraries and technologies, including:
This app also utilized the Spotify authentication flow, including management of access and refresh tokens.
Spotify Music Visualization App