My main job at imbellus was building an editor that allowed our task designers to construct tasks without needing regular engineering or graphic design support. Originally I was hired as a contractor to build out the initial version of the editor, but as the editor became a regular tool in our task creation process I was hired to continue developing and maintaining it. This role involved a lot of work directly with my users, adapting the tool to their needs as they evolved. In addition, I had to adapt some of the infrustructure on our main client app to support this.

Some of the features I’m most proud of included:

• A system for automatically generating both front-end and back-end code, and emulating the back end in editor - allowing the designers to build the task and test it in a manner that matched the production version
• A visual programming language with robust error detection, allowing the designers to construct logic without needing to write code directly for most situations
• A typescript compiler & editor, allowing the designers to write custom code where needed, and have it validated against the rest of the task
• The capacity to live-reload the preview, including 3D assets, with any updates to logic as you build
• Integrating a Unity3D based preview directly into the editor

In addition to the editor, I was occasionally pulled in to support efforts in other areas. For example:

• Adjusting the structure of communication between Javascript & Unity, to move the “source of truth” into our Javascript client - and have the Unity 3d view just match the state it is given.
• Help migrate from our tasks being embeded into a Unity desktop application, to our tasks running in a browser using Unity’s WebGL builds.
• Build a back-end execution layer in our Python server that could run Javascript code generated by our editor, and store it’s state in memory isolated from other instances - allowing for real-time, stateful task evaluation on the backend while a client is interacting with it on the front end.
• Develop a web-component based “shell” for our next-generation tasks - with each task acting as a micro-frontend, and the global state & connection to the server managed by the shell.