David Geffen Galleries at LACMA — Computational Geometry in Design Development

In Brief

The David Geffen Galleries presented an unusual computational challenge: a fluid, irregular building geometry that did not fit conventional rectilinear modeling logic. During early design development, I used Python to explore Delaunay and Voronoi-based systems, testing ways to organize and rationalize the evolving geometry.

The Geometry Problem

Designed by Atelier Peter Zumthor in collaboration with SOM, the galleries form an elevated, irregular building that crosses Wilshire Boulevard. Its long curves and changing boundaries made the geometry difficult to describe with a conventional orthogonal grid alone.

During design development, computational methods made it possible to test relationships among points, boundaries, surfaces, and subdivisions as the geometry changed. Delaunay and Voronoi were explored as tools for those studies; these explorations should not be read as a claim that either became the final architectural or façade system.

View the Exhibition Level plan at full size

Delaunay Exploration

Delaunay triangulation connects a set of points into a network of triangles. When point locations change, the relationships between them can update, making the method useful for testing how an irregular field might be subdivided.

I used Python to generate and test Delaunay strategies during early design development. Iterating the point distribution and mesh density made it faster to compare geometric relationships and produce studies for the evolving form.

Interactive footprint study

This explorer uses a representative, simplified footprint and a seeded set of points. Adjust the density to see the triangulation respond. It demonstrates the input, algorithmic relationship, and iteration process; it is not a survey-accurate building plan or a record of a final project model.

Voronoi Exploration

A Voronoi system divides space into regions around seed points: every location in a region is closer to its seed than to the others. Changing the seed points changes the resulting cells, which makes the method useful for exploring cellular subdivisions and panelization patterns.

Voronoi-based geometry was tested as part of early geometry and panelization exploration. The studies helped examine alternative ways to organize a changing surface; they are not presented as the final façade or roof construction system.

My Role

BIM / Design Technology / Computational Design

• Developed Python-based computational studies during early design development.
• Explored Delaunay triangulation strategies for irregular geometry.
• Tested Voronoi-based spatial and panelization systems.
• Supported iterative geometry development and connected computational exploration with BIM and design-technology workflows.

From Computation to BIM

The value of the code was not the algorithm itself. It was the ability to respond quickly as the design changed. Python studies helped compare alternatives and communicate changing geometric relationships within the broader BIM and design-technology workflow. The drawings below explain that transition from geometric exploration to coordinated building systems; they do not represent the original project models or a claim that I developed every system shown.

View the exploded systems illustration at full size

Roof framing study

This retained comparison switches between a roof model render and its wireframe view. It helps show how the flowing outline meets a more regular framing field. The images are explanatory portfolio views, not an interactive 3D model or archival BIM documentation.

Geometry, framing, and systems together

The combined illustration brings the roof outline, framing, and selected service layers into one view. It conveys the coordination problem created by a changing, irregular perimeter without claiming that this diagram records the final constructed arrangement.

View the roof and systems illustration at full size

The Built Project’s Technical Context

The completed galleries are a substantial concrete structure with long cantilevers and seismic isolation. SOM describes 40 triple-friction-pendulum base isolators and cantilevers reaching up to 80 feet. The architectural concrete required more than 100 distinct in-place pours, sequenced to manage consistency and curing. These verified construction facts provide context for the project’s technical complexity; they are separate from my computational-study contribution.

SOM lists 2026 as the completion year and reported that the galleries opened to the public on May 4, 2026. Project area figures vary across project-stage sources; I have left area out until it can be confirmed against a final official record.

Project Attribution

Building design: Atelier Peter Zumthor, in collaboration with SOM. SOM identifies its role as architect of record and structural engineer. My contribution described here is early-stage computational design development using Python, including Delaunay and Voronoi explorations in the broader BIM and design-technology workflow. I did not design the building’s architecture.

The hero and selected technical visuals were supplied for this portfolio case study. The plan, exploded axonometric, and roof coordination illustration are explanatory reconstructions based on project references; they are not archival project drawings, original BIM screenshots, or construction documents.

What this project demonstrates

• Using computation during real design development.
• Applying Python to explore complex architectural geometry.
• Turning irregular form into systems that can be tested and iterated.
• Connecting computational exploration with BIM and project development.
• Using algorithms as design tools rather than visual effects.

References

• SOM — David Geffen Galleries project facts, design and engineering
• SOM — galleries complete and public opening date
• SOM — project collaboration and architect-of-record role
• LACMA — concrete construction and project-team coordination