Justin builds useful AI systems.
I specialize in software that bridges creativity and technology: node-based AI IDEs, 3D tools, ML models, and real applications with real users.
AI IDEs
3D tools
ML models
Full-stack
Clean architecture. Practical tooling. Shipped work.
- Focus
- AI
- Stack
- Full
- Work
- Tools
- Status
- Open
Snapshot
Built for creative AI workflows.
A concise read for recruiters and employers: what I build, where I fit, and how to reach me.
Work
Projects.
Selected software and tools I've built.
01
3D app
Roka3D
An image-to-3D and text-to-3D webapp for making miniature figurines for TTRPGs.
Docker
Deno
ComfyUI
Hunyuan 3D
02
AI IDE
Neu
A node-based IDE for AI-enabled apps with ComfyUI integration, custom nodes, community features, and a built-in web server.
Java
ffmpeg
OpenAI
Anthropic
03
Workspace
NeuOS
An experimental OS-like workspace for executing GUI apps made in Neu.
Java
ffmpeg
GUI apps
Open source
Open Source.
A compact set of MIT-licensed repositories around agent tooling, Swing UI systems, and console app infrastructure.
A desktop coding agent with a terminal-style UI, multi-provider LLM support, file tools, shell commands, Maven workflow, web browsing, Git review, and Cursor-style skills.
A modern Java Swing Look and Feel inspired by shadcn, with light/dark themes, styled controls, demos for dialogs, kanban, notes, image cropping, and node graphs.
A Java 17 Swing console UI toolkit for keyboard-driven desktop apps, with character-cell rendering, widgets, file drops, dialogs, diff review, and optional AI-assisted input.
Model work
LoRA & model work.
Fine-tuned models and LoRAs for image editing, generation, and creative production.
Improves Qwen Image Edit clothing transfer while preserving body shape and accurate garments.
Turns a person image into a clean outfit extraction on a white background.
Fixes photos of screens by converting them into cleaner screenshot-style outputs.
Helps Qwen Image Edit preserve more accurate pose transfer.
A Z-Image Turbo LoRA for making images of 3D models and tabletop RPG minifigures.
A pixel-art LoRA for Game Boy Advance and DS-era Pokemon trainer generation.
Capabilities
Build range.
The core areas I can contribute to across AI product and creative tooling teams.
model
LoRA training, image editing, generation workflows
tools
node-based IDEs, ComfyUI integrations, custom nodes
app
full-stack webapps, intuitive servers, native GUI apps
media
image-to-3D, text-to-3D, ffmpeg, creative pipelines
ship
clean architecture, practical tooling, minimal dependencies
AI application tooling
Interfaces for building, testing, and running AI-enabled software.
Creative ML systems
LoRAs, image workflows, model editing, and generation pipelines.
3D and media tools
Webapps and pipelines for 3D models, miniatures, and GUI workflows.
Lean engineering
Deno, Tailwind, minimal dependencies, and direct product feedback.