How I work across design and code

Material UI already covers most of what a data-dense product needs: the question is what can we build with our current resources. On the design system I went through every customized component the team had in use and asked whoever made it to justify going past the defaults. Almost none of them could.

This allowed me to build for the few exceptions and get people to stop customizing components.

Our engineering team was in Poland. I needed to hand off work and have them going while I was asleep.

I typed the theme file in TypeScript, so the editor offers small, medium, large instead of raw values, and wrote the documentation in Docusaurus with examples pulled from React.

Status icons on each Figma page showed what was done and what was open.

If the query pulls the wrong column, a chart can still look fine.

The first version of the tariff dashboard connected the database variables wrong and returned an incorrect result.

I didn’t diagnose the visualization: I found the error in the code. When I use a coding model on a real dataset I feed it one step at a time and read what it generates.

Operators schedule cleaning across hundreds of furnaces. AI models were implemented to give recommendations, but operators had the final decisions.

I separated model output from measured readings on the dashboard, so an operator could always tell if a person or an AI was making the judgment call. The schedule stayed editable to avoid AI making costly decisions.