Applied AI
Efficient models, structured decisions, and local inference.
Applied AI · IoT · Robotics
Danger Room
Practical intelligence for connected systems.
Danger Room explores the intersection of artificial intelligence, connected devices, and robotics through practical experiments and reproducible engineering. This is a home for public models, reusable pipelines, and documentation that makes the work useful beyond a demo.
Focus areas
IoT & edge intelligence
Sensor-driven workflows and intelligence close to connected devices.
Robotics
Perception, simulation, and the connection between software decisions and physical action.
Featured release
MadStickArt v0.8.0
Clean stick-figure storyboard scenes at 956 × 400 pixels (2.39:1). Structured scene decisions and procedural composition drive a compact trained neural decoder for rough storyboarding and shot planning on Apple silicon.
Apache 2.0 · Local inference · Runnable examples · Documented benchmarks
Try the model online → Explore the latest stable release → Versioned model entry → Latest stable on main →
The separate original MadStickArt v0.1.0 repository remains available. Pin a version tag for reproducible downloads.
Built to be inspected and reused
Public releases prioritize clear licenses, versioned artifacts, reproducible workflows, and evidence behind performance claims. Each release documents its supported use, evaluation conditions, and limitations.
Technical questions, reproducibility reports, and ideas for future work are welcome in the MadStickArt Community tab.