Edge AI · Jetson Orin, Orin Super, Thor
NetBytes Jetson Orin edge AI solutions
Industrial-grade vision AI, generative AI, robotics, machine vision, medical imaging and intelligent transportation systems on NVIDIA Jetson — engineered, optimised and supported in India.
- Orin / Orin Super — Next-gen edge AI compute
- JetPack 6.0 — CUDA, TensorRT, cuDNN
- DeepStream — Vision pipelines
- Isaac + ROS 2 — Robotics stack
- Industrial build — Enclosures and cooling
About the platform
Edge AI compute, engineered into something deployable
Industrial-grade edge AI, vision AI, generative AI, robotics, machine vision, medical imaging and intelligent transportation systems built on NVIDIA Jetson Orin, Orin Super and Jetson Thor.
A Jetson module is not a product. What makes it deployable is the enclosure, the cooling, the I/O, the OS image, the update path and someone to call — which is the part Spectra engineers, optimises and supports.
The module choice follows the workload: Orin Nano and NX for light to mid workloads, Orin Super for high-density AI vision, and Jetson Thor for robotics and multimodal AI.
Family comparison
Which module, and why
Compute, memory and application class — the three inputs that decide the module before anything else is specified.
- Jetson Orin NX / Nano — AI performance: Up to 100 TOPS. Memory: 8–16GB. Recommended for: Vision AI, compact robotics, light edge inference
- Jetson Orin Super — AI performance: Up to 200 TOPS. Memory: 32GB. Recommended for: High-density multi-stream analytics, machine vision, robotics
- Jetson Thor — AI performance: Up to 275 TOPS. Memory: High-capacity unified memory. Recommended for: Advanced robotics, humanoid robots, generative and multimodal inference
Orin Super offers up to 70% more AI performance than Orin NX and Nano. Thor adds transformer engine acceleration for vision-language and multimodal models.
Product lines
Five lines these get built into, and a sixth for everything else
- Edge Vision AI Box — On-edge video analysis, event alerts, cloud-edge orchestration, remote management and AI-enhanced CCTV applications.
- Edge Generative AI — On-device inference for vision-language models, zero-shot detection, local multimodal models and real-time agent workflows.
- Machine Vision Controller — High-accuracy inference for inspection, defect detection, robotic arms and manufacturing AI workflows.
- V2X Roadside MEC — Edge compute for smart mobility — vehicle-road-cloud fusion, roadside perception, V2X communication and traffic analytics.
- Medical Imaging Controllers — AI-accelerated CT, DR and MRI inference, endoscopy imaging, ultrasound analysis and surgical robotics controllers.
- Custom builds — I/O, storage, enclosure and OS image configured to the deployment rather than to a catalogue entry.
Why NetBytes
The work between a dev kit and a deployment
A Jetson development kit proves a model works. Getting that model running unattended, in a cabinet, on a plant floor or at a roadside, for three years, is a different exercise entirely.
That gap is where the engineering sits: enclosure and thermal design for the environment, an update path that does not require a site visit, and a pipeline someone will maintain after the pilot ends.
- Full hardware and software integration, delivered as one system
- Industrial-grade enclosures and cooling, specified for the environment
- OTA updates and device fleet management
- On-site deployment and maintenance
- Custom AI pipeline development, from training through to deployment
- toolkit for training
- for optimisation
- and Isaac for deployment
- fleet updates
Applications
Where these are deployed
- Real-time vision AI — object detection, tracking and behaviour analysis
- Edge generative AI — vision-language models powering intelligent agents
- Robotics — autonomous mobile robots, manipulators and industrial arms
- Machine vision — inspection, metrology, quality assurance and defect detection
- Smart transportation — V2X, traffic analytics and edge MEC
- Medical imaging — CT, MRI and ultrasound inference at the edge
Jetson — FAQ
What integrators ask first
- Which Jetson module should I choose? — Answer: Orin Nano or NX for light and mid workloads, Orin Super for high-density AI vision, and Jetson Thor for robotics and multimodal AI.
- Do you support end-to-end AI pipeline development? — Answer: Yes — from model training with the TAO Toolkit, through optimisation with TensorRT, to deployment on DeepStream or Isaac.
- Can these integrate with existing CCTV or robotic systems? — Answer: Yes. The devices support RTSP, ONVIF, industrial fieldbus, ROS 2 and custom APIs.
- Do you offer cloud integration? — Answer: Yes — integration with AWS, Azure, or a custom private edge cluster.
Size it first
Work the numbers before you specify
The calculators run the same arithmetic our engineers do, so you can arrive with a starting configuration rather than a blank page.
- Video analytics GPU sizing — Streams, models and frame rate to a GPU count that holds under real load.
- Face recognition GPU sizing — Faces per second and template database size to GPU and memory.
- All sizing tools — Six calculators running the same arithmetic our engineers do.
Related range
Related platforms
Where the edge meets the rack.
- Rack GPU AI servers — Training and heavy inference behind the edge estate.
- Compact PCs — Fanless and compact systems for kiosks, signage and equipment rooms.
- SNS rack storage — Where the data the edge produces is retained.