A live, large-scale VR product that had outgrown the architecture it was built on. I came in as a senior developer to re-design the core systems so it could keep adding features without each one costing more than the last.
The product was already live and already large, which is the hardest version of an architecture problem. You can't stop, you can't start again, and every change has to be shippable on its own. The symptom was the usual one: feature velocity falling, each addition touching more of the codebase than the one before it.
The re-design targeted sustainable feature scaling rather than elegance for its own sake. The measure I cared about was simple. How many files does a new feature have to open? Bringing that number down meant establishing clear ownership of state and removing the implicit coupling that had accumulated between systems shipped at different times.
The second piece was an interaction system on a ScriptableObject architecture, with custom inspectors and editor windows, so designers could author complex behaviour without engineering support. This is the same conviction that later became GripCraft and ReactiveVars. Highstreet is where I learned how much of a VR product's schedule really goes on waiting for a programmer to wire something up.
I also helped recruit and onboard the core engineering team, running technical interviews and contributing to hiring decisions. Part of fixing an architecture is making sure the people who inherit it understand why it is shaped that way.
I don't hold capture rights to this title, so there are no screenshots here. The product is publicly available on the Meta Quest Store if you would like to look at it directly.