Janet Echelman — Computational Design for Tensioned Net Sculpture

In Brief

Large-scale tensioned net sculptures need geometry that can remain expressive while responding to structural constraints. At SOM, I worked on computational systems for framing and net geometry, using Python and parametric workflows to connect anchor points and support design iteration.

I also built an early voice-driven Revit experiment that allowed geometry calculations to be triggered through Amazon Alexa.

My Role

Parametric Design Technologist

• Developed scripted geometry systems for structural framing logic
• Implemented point-to-point connection algorithms for net geometry
• Modeled layered net structures with parametric control
• Supported structural rationalization through computational workflows
• Built an experimental Amazon Alexa → ngrok → Revit API pipeline to trigger geometry calculations

The Engineering Problem

Dream Catcher, by artist Janet Echelman, is suspended between two hotel buildings in West Hollywood, California. SOM collaborated on its engineering and integration with the architecture. SOM's account of the collaboration describes adapting the net geometry to the site's contours.

A tensioned net is a dynamic structure. Anchor locations, spans, rope geometry, attachment conditions, and environmental loads interact: a change at the boundary can change the form across the network. The engineering team's task was to keep material behavior and structural logic aligned with the artist's intent. My contribution focused on the computational geometry workflows described here.

Computational Geometry

The challenge was not simply drawing the net. It was maintaining a geometric system that could respond when anchor points, spans, or structural conditions changed. Point-to-point connection logic and layered parametric modeling made those relationships explicit, supporting structural rationalization and updated geometry.

Geometry workflow

1. 01

   Anchor points

   Locate the model's connection conditions.

2. 02

   Connection logic

   Calculate point-to-point relationships and net topology.

3. 03

   Geometric calculation

   Update spans and layered net geometry for review against structural and fabrication constraints.

4. 04

   Updated model

   Carry revised relationships into the parametric model for another design iteration.

Dream Catcher — Interactive Weave Explorer

Interactive portfolio reconstruction inspired by the geometry of tensioned net systems. This browser study illustrates connection logic and spring relaxation; it is not an original project model or an engineering analysis of Dream Catcher.

Select Generate Points, then Connect Threads. Adjust Radius to explore local connections and wider spans. Points applies on the next generation; Relaxation applies on the next connection. Toggle the boundary or nodes to inspect the network, or Reset to clear it.

An Early Voice-Driven Revit Experiment

Voice served as an interface to a controlled geometry workflow. I built a pipeline from Amazon Alexa, through ngrok, to the Revit API. A voice command triggered the point-to-point connection algorithm and recalculated net geometry in Revit in real time.

The important piece was the API and geometry pipeline: the command initiated an actual model calculation. This was an experiment in voice interaction, with a different implementation from today's AI-agent systems.

From Computation to Engineering

My scripts connected framing logic, layered net geometry, and the digital model so that changing relationships could be explored through computation. This work sat between art, architecture, and engineering, where geometric choices had to be reviewed alongside attachment conditions and material behavior. The computational model supported that exchange; engineering decisions remained with the project team.

What this project demonstrates

• Using code to work with non-standard structural geometry
• Translating anchor conditions into parametric net relationships
• Connecting computational design with Revit workflows
• Testing voice interaction with design software before modern AI-agent workflows
• Supporting exchange between artistic intent, engineering constraints, and the digital model

Connection to Current Work

The recurring pattern is user input entering a controlled system, computation translating that intent into geometry, and Revit holding the model result. The voice experiment and my current AI work use different technologies, but both depend on a defined path from input to a Revit action.

See AI → Revit: Prototype to Production Systems for the later work, or Poly International Plaza for another application of parametric structural geometry.

Credits & References

Portfolio case study: Danny Bentley, Parametric Design Technologist at SOM. Artwork: Janet Echelman. Engineering collaborator: SOM. Dream Catcher is the primary project example; the responsibilities above describe my existing account of computational work at SOM, rather than the full team's engineering scope.

• SOM — Creative Construction: How Artists and Engineers Collaborate
• SOM — Art and Engineering: A History of Collaboration
• SOM — Award for Collaborative Artwork in West Hollywood