01 / NETWORK SIMULATION
Follow a packet. Understand the system.
TypeScript · Discrete-event simulationInteractive
SIMULATION / DATA / EXPLORABLE WORLDS
Software you can step inside.
Six projects, built to be explored.
Open an application or inspect the engineering behind it.
02 / GEOSPATIAL / 3D
A city you can read at street level.
Three.js · BlenderInteractive
03 / GEOSPATIAL / SCENARIOS
One real city. More possible futures.
Three.js · LiDAR terrainInteractive
04 / CREATIVE GRAPHICS
Step beyond the edge of a painting.
Three.js · Authored geometryInteractive
05 / NUMERICAL SIMULATION
The same laws. Radically different futures.
JavaScript · Web WorkersInteractive
06 / DATA INTERFACES
A race, from several angles.
React · TypeScriptInteractive
THE INTERACTIVE WORKBENCH
Fabric Reality Lab
Explore the actual application.
One project at a time, right here.
NETWORK SIMULATION
Follow a packet. Understand the system.
Start the guided tour, then compare architectures and inspect a packet.
Opens here · Loads only when you launch itChoose an image to launch, or a project title for technical details. 3D views require WebGL. Each project opens on its own; closing it stops its rendering and background work.
WHAT IT DOES
An interactive HPC and AI network simulator. Compare fabric architectures, inspect queues, and trace how a collective workload moves through the network.
INSIDE THE BUILD
01
128 explicit ranks and 32 access switches across three modeled fabric architectures.
02
Deterministic packet scheduling, bounded queues, and full-duplex link serialization.
03
Paired architecture replay, round waterfalls, traffic matrices, and latency distributions.
Model boundaries & provenance +
A reduced, explainable network model—not a vendor performance benchmark. Model assumptions are exposed in the lab. Each project is deployed independently from its own GitHub repository.
ENGINEERING NOTES
The interesting part
is how it behaves.
Simulation, visualization, and data-heavy interfaces share a problem: the user needs to see what a system is doing—and understand what the model can actually tell them.
These projects make that behavior inspectable through controls, comparisons, and explicit model boundaries. The source links go straight to the implementation.
Project descriptions are grounded in implementation source. Each application includes its own model boundaries; each live project is maintained in its own repository.