In this project, I worked with a unique constraint: I only had access to the generated website output—HTML, CSS, and JavaScript—produced by a proprietary builder developed by a third-party company. The goal was to support teachers in creating online exams, but the builder's output was inflexible, repetitive, and often required manual fixes.
To address this, I developed a custom automation workflow using Python and Vanilla JavaScript. The core of my solution was a set of Python scripts that automatically identified and resolved common issues in the generated code. These included accessibility bugs, layout inconsistencies, and missing assets—things that previously had to be handled manually for every single exam.
I deliberately chose to work entirely from the terminal using Neovim to challenge myself and improve my fluency in low-overhead environments. Most of my workflow revolved around quickly iterating on scripts and doing targeted manual edits directly in the raw output when needed.
The project had a strong human component as well. Teachers had the freedom to request custom features or layouts. I acted as the bridge between their needs and the rigid system, modifying the output code to match their vision—often with very little standardization or consistency in the inputs.
Although I attempted to automate testing, the nature of the system made this difficult. Each output could differ significantly in structure and content, and there was minimal standardization. This made manual QA an essential part of the process to ensure that the final product was reliable and production-ready.
