FlashRT is a high-performance realtime inference engine for small-batch, latency-sensitive AI workloads. The flagship integration is production VLA control for Pi0, Pi0.5, GROOT N1.6, and Pi0-FAST. Also support llm e.g, qwen3.6-27B

AI agents running research on single-GPU nanochat training automatically for Jetson Thor

Lightweight Rust TUI system monitor for NVIDIA Jetson and Linux. Track CPU, RAM, GPU temps, network, disk I/O, and processes.

THOR — Native macOS control center for NVIDIA Jetson devices. Manage power, Docker, ROS2, ANIMA AI modules, cameras, GPIO, and more. No terminal needed.

Real-world voice AI model benchmarks on NVIDIA Jetson Thor (JetPack 7.0, Blackwell GPU, 122GB unified RAM). Comparing STT, TTS and omni models for low-latency voice assistants.

Docker packaging to run antirez/ds4 (DwarfStar / DeepSeek-V4-Flash) on NVIDIA aarch64 unified-memory boxes - tested on Jetson AGX Thor (sm_110 , CUDA 13), persistent disk KV cache.

AI drawing robot: a 4-DOF Waveshare RoArm-M2-S draws with a felt pen on a vertical board, driven from a Jetson AGX Thor — backlash-compensated IK, closed-loop encoder verification, and wrist-cam visual servoing.