Nexus Emerging Technology

Build, train and run AI on your own hardware.

Explore Nexus tools for local inference, model training, compression, optimisation and productivity.

Explore the suite Try Nexus Office

Nexus Office — browser trial

Nexus Office workspace — home screen showing Writer, Grid, Dashboard Studio, NexusPoint, Paint, and Compatibility app cards

A coherent suite for local AI work.

Each tool addresses a specific stage of the development workflow.

AI Development

Nexus AI

Active

Local inference and model interaction.

v—View Project

Random Initialization Foundry

Active

Training models from random initialisation using a desktop application with a PyTorch backend.

v—View Project

Compression & Recovery Studio

Active

Model compression and recovery workflows.

v—View Project

Adapter Optimization Studio

Active

Adapter training and optimisation.

v—View Project

Benchmark Studio

Active

Measuring and comparing supported execution settings.

v—View Project

Productivity

Nexus Office

Active

Browser-accessible productivity tools, with additional capabilities available through the local installation.

v—Try in Browser

Browser trial available

Nexus Office runs as a self-contained HTML application. No account required for the core trial.

Open trial

Experimental Computing

Note

Projects in this group are at an early or experimental stage. Maturity is identified on each project page.

Nexus OS

Experimental

Experimental operating system project.

v—View Project

Virtual Machine Projects

Experimental

Virtual-machine tooling and imaging/flashing utilities.

v—View Project

Evaluate before you commit.

Nexus Office is available as a self-contained browser application — no account required, no installation needed for the core trial.

Self-contained HTML interfaces make browser evaluation convenient. Older-format conversion and PKI document signing require the local installation. Not all features run entirely in the browser.

Open Nexus Office View repository

rdunbar01.github.io/Nexus-Office/

Measurements, not marketing.

The Research & Benchmarks section contains engineering notes and benchmark results with full methodology — model, hardware, operating system, workload, precision and measurement method. Training throughput, prompt-processing speed and generated-token speed are reported separately. User-recorded results are labelled as such.

View Research & Benchmarks

Methodology fields

Model identifier and revision
Hardware (device, VRAM, driver)
Operating system and runtime version
Workload type and sequence length
Precision and quantisation settings
Measurement method and tool
Training throughput (tokens/s)
Prompt-processing speed (tokens/s)
Generated-token speed (tokens/s)

About the project

Nexus Emerging Technology is an independent software project by Richard Dunbar, focused on building practical tools for local AI development. The suite addresses the full workflow from model training and compression through inference and productivity, with an emphasis on running workloads on hardware you control.

github.com/RDunbar01