Getting Started with Jetson Orin Nano Super Developer Kit: First Boot to First AI Model

The Jetson Orin Nano Super Developer Kit is the most accessible entry point to NVIDIA-powered edge AI development. At $249, it delivers 67 TOPS of AI performance — enough to run generative AI models locally. But getting from unboxing to running your first AI model requires a few careful steps.

This guide walks you through the complete setup process: hardware preparation, OS installation, software configuration, and running your first AI inference model.

What’s in the Box

• Jetson Orin Nano 8GB module (pre-mounted on carrier board)
• Reference carrier board with all I/O exposed
• Power supply (DC, 5.5mm × 2.5mm jack)
• Quick start guide

You’ll also need: A microSD card (64GB+ recommended, UHS-I or faster), a USB keyboard and mouse, a DisplayPort monitor (or DP-to-HDMI adapter), and an Ethernet cable for initial setup.

Step 1: Prepare the microSD Card

The Jetson Orin Nano boots from a microSD card (or NVMe SSD for better performance). Start with the microSD approach for initial setup.

1. Download the latest JetPack SDK image from developer.nvidia.com/embedded/jetpack
2. Flash the image to your microSD card using Etcher (balenaetcher.io) or the dd command on Linux
3. Insert the flashed microSD card into the slot on the underside of the developer kit

Step 2: Connect and Power On

1. Connect your DisplayPort monitor
2. Connect USB keyboard and mouse to any of the 6 USB 3.2 ports
3. Connect Ethernet for internet access
4. Connect the power supply — the system boots automatically

The first boot takes 2-3 minutes. You’ll see the NVIDIA logo, then the Ubuntu desktop setup wizard.

Step 3: Initial Configuration

1. Accept the NVIDIA license agreement
2. Choose your language, timezone, and keyboard layout
3. Create your username and password
4. Let the system complete its initial configuration (this may take several minutes)

Once you reach the Ubuntu desktop, open a terminal and verify your setup:

# Check JetPack version
cat /etc/nv_tegra_release

# Verify GPU is recognized
nvidia-smi

# Check CUDA installation
nvcc --version

Step 4: Update the System

sudo apt update && sudo apt upgrade -y

This ensures you have the latest security patches and driver updates.

Step 5: Install AI Frameworks

JetPack comes with CUDA, cuDNN, and TensorRT pre-installed. For Python-based AI development, install the Jetson-optimized versions:

# Install pip if not present
sudo apt install python3-pip -y

# Install PyTorch (Jetson-optimized wheel)
# Check developer.nvidia.com/embedded for the latest URL
pip3 install --upgrade torch torchvision

# Install ONNX Runtime with TensorRT provider
pip3 install onnxruntime-gpu

Step 6: Run Your First AI Model

Let’s run a simple image classification model using TensorRT for optimized inference:

# Clone NVIDIA's Jetson inference examples
git clone --recursive https://github.com/dusty-nv/jetson-inference
cd jetson-inference
mkdir build && cd build
cmake ..
make -j$(nproc)
sudo make install

# Run image classification on a sample image
imagenet --input images/orange_0.jpg --output output.jpg

This downloads a pre-trained image classification model, optimizes it with TensorRT, and runs inference on a sample image. You should see the classification result in your terminal and the annotated image saved as output.jpg.

Step 7: Try Real-Time Camera Inference

Connect a USB camera and run real-time object detection:

detectnet /dev/video0

This launches a real-time object detection stream using the SSD-Mobilenet model, optimized with TensorRT. You should see bounding boxes around detected objects in the camera feed.

Optional: Upgrade to NVMe Storage

For production workloads, the microSD card becomes a bottleneck. Install an NVMe SSD in the M.2 Key M slot for dramatically faster storage:

1. Power off the developer kit
2. Install an M.2 2280 NVMe SSD in the Key M slot
3. Boot from microSD, then use the NVIDIA SDK Manager or dd to clone your system to the NVMe drive
4. Update the boot configuration to boot from NVMe

Next Steps

Now that your Orin Nano is running, explore these resources:

• NVIDIA Jetson AI Lab: Pre-built containers for LLMs, VLMs, and generative AI on Jetson
• TensorRT: Optimize your custom models for maximum inference throughput
• DeepStream SDK: Build multi-stream video analytics pipelines
• Isaac ROS: Robotics-specific AI acceleration packages

Ready to take your edge AI project further? Contact us for module selection, carrier board recommendations, and production deployment guidance.