Show HN: Optimized runtimes for three VLAs on Jetson Thor

Category: ai-ml

Tags: vla, jetson, cuda, tensorrt, robotics, edge-inference, model-serving

Score: 6.3/10 (Innovation: 6, Technical: 7, Documentation: 7, Utility: 5)

vla-edge provides optimized CUDA kernels and serving infrastructure for three Vision-Language-Action (VLA) models on the NVIDIA Jetson AGX Thor edge platform, achieving 2-5x speedups over PyTorch and TensorRT baselines. It includes deployment recipes, a shared HTTP API for robot control, and pre-compiled model bundles, making VLA robotics accessible on embedded hardware.

Target audience: robotics engineers, ML engineers, embedded AI developers

Repository: https://github.com/Agents2AgentsAI/vla-edge · Python · Apache-2.0 · 9 stars

View on Hacker News