This past weekend at the AEC Tech Hackathon in NYC, I pitched an idea which drew together nine talented software developers and computational designers toward building UpZone, then I led that team in building the product over the course of 24 hours.
Role
Product Lead
Competition Winner
Best Breakout Hack, AEC Tech Hackathon 2024
Hackers' Favorite, AEC Tech Hackathon 2024
What is UpZone?
UpZone tackles the massive task that architects currently face of sifting through thousand-page zoning regulations for their particular site before even beginning design.
It is a one-click solution to identifying any NYC lot's maximum buildable zoning volume and creating a volume that exports directly into Rhino so that architects can hit the ground running with design work.
How Does It Work?
UpZone utilizes LLMs to analyze zoning documents and search city planning websites, then generates site-specific 3D volumes.
Team Process
I organized our team of nine into four subgroups, each one tasked with a different segment of the product build.
- Group 1: LLM Ops
- This team developed two parallel processes which analyze the 4,000-page NYC Zoning Resolution using both Gemini and ChatGPT. The LLMs process the Resolution and output values that establish site geometry. These values include maximum building height, sky exposure plane, front / rear / side setbacks, and more.
- Group 2: GIS Data
- This pair gathered GIS data from city planning websites and generated geoJSONs for the UpZone base map.
- Group 3: Grasshopper Geometry
- This pair built out a grasshopper script that assigns the values that Group 1 pulls from the Resolution docs to variables of a 3D volume.
- Group 4: Shapediver App
- This group gathered all of the outputs from Groups 1-3 and compiled them into a Grasshopper script for a Shapediver-based application.
- UpZone works on both desktop and mobile.