I wanted to explore UI ideas for ETL processes. Can we make complex data transformations as easy to build+troubleshoot as spreadsheets? What if we had tools to quickly diagnose+fix common issues in data pipelines?

To answer those questions, I built an interactive prototype in React that parses JS formulas as they’re edited.

Built using Hugo, Tailwind CSS, and an obsessive emphasis on performance. You can read more about it here. It has a dark mode using the new prefers-color-scheme media query and the Solarized colour scheme, because I love Solarized Dark.

This started out small, as a tiny C# and WPF urban planning tool to help me visualize the impact of setback and lot size requirements on individual properties. Eventually I rewrote it as a web app using the HTML5 Canvas element, and after that I realized that I might as well scale it to entire neighbourhoods…

I built an interactive map of every property in Vancouver using Mapbox, Eric Fischer’s remarkable Tippecanoe, and a Heroku app to serve the map tiles.

Be warned, there is a bug in Tippecanoe that makes the map look a little wonky at lower zoom levels and the free Heroku tier takes ~10s to serve map data.

This has two parts:

1. An Angular front-end that generates personalized letters in support of housing projects
2. A Serverless Framework back-end that uses Slack and multiple AWS services (Lambda, SES, S3, Athena) to review+approve+send+analyze emails

I wrote the front-end and the incomparable Roman Lisagor wrote the initial version of the back end, but these days I’m hacking away at both. More about the project can be found here.

This little thing scrapes the City of Vancouver’s rezoning application page then notifies of any changes via Slack. It’s an .NET Core Azure Function written in C# using Azure Table+Blob storage.