CTai LABS Application Kits
Technical guides, decision frameworks, videos, case studies, manuals, and reference articles for engineers taking ai from prototype to production at the Edge.
Turn Physical ai Capabilities into a Practical Starting Point
CTai LABS kits combine application-ready ai, Connect Tech Edge compute, supporting hardware, and engineering guidance to help teams evaluate advanced capabilities without assembling every system layer independently.
Each kit provides a defined starting point for testing a real use case on real hardware. CTai LABS configures the ai environment, connects the required sensors and software, optimizes the application for the target platform, and helps customers build toward their deployment requirements.
Your team keeps its data, proprietary process knowledge, application, system design, and project IP throughout the engagement.
Choose Your Development Path
Choose the capability closest to your application. Select “View Kit” to jump to its detailed section, block diagram, Connect Tech product list, and next steps.
Scene Analyzer Agent Kit
Turn live camera streams and recorded video into locally processed, searchable scene intelligence. Generate summaries, configure alerts, retrieve relevant clips, and ask natural-language questions grounded in stored video evidence.
On-device video understanding
Configurable alerts
Searchable local evidence
Natural-language Q&A
3D Spatial Intelligence Kit
Transform a single monocular camera stream into real-time depth information, colorized point clouds, and structured 3D scene understanding. Give robots usable information about objects, layouts, reachability, clearance, and free space.
Single-camera 3D reconstruction
Spatial reasoning
ROS2-compatible outputs
Local Edge processing
Physical ai Evaluation Kit
Evaluate a complete Real2Sim-Sim2Real workflow connecting live sensor data, digital twins, foundation models, robotics, and Edge deployment. Explore how physical environments can inform simulation and how validated models and policies can return to physical systems.
Real2Sim digital twins
Multi-view 3D tracking
Scene understanding
Sim2Real evaluation
ScrapGuard™ Hardware Kit
Identify sealed pressurized cylinders in scrap-processing streams before they reach a downstream shredder. ScrapGuard™ combines rugged Edge compute, industrial vision, local inference, and a PLC-facing result in one site-configured package.
Industrial hazard detection
Three-stage perception
Local Edge inference
PLC-facing output
How Our Kits Work
From concept to development in three simple steps.
Choose a Starting Point
Select the kit that best matches your application, operating environment, sensors, and technical goals.
Validate Your System
Evaluate the application using configured hardware and software, representative data, and defined performance criteria.
Build Towards Deployment
Work with CTai LABS to optimize and customize the application’s models, sensors, hardware configuration, interfaces, and validation approach.
Kit configurations vary based on application requirements, sensor interfaces, performance targets, operating environment, and component availability. CTai LABS confirms the final hardware, software, and support scope with each customer.
Not sure which kit fits?
Bring us an idea, sample video, dataset, sensor requirement, existing model, prototype, or complete installation brief. CTai LABS can recommend the strongest starting point and identify where customization may be needed.
Scene Analyzer Agent Kit
Turn Video into Searchable Scene Intelligence
The Scene Analyzer Agent Kit turns live camera streams and recorded video into structured summaries, configurable alerts, searchable clips, and natural-language answers. Video, metadata, and generated insights can remain on the Edge system, without relying on cloud connectivity for the core workflow.
Developed by CTai LABS, Scene Analyzer Agent combines video ingestion, Vision-Language Model (VLM) scene understanding, NVIDIA® Nemotron-based reasoning, local storage, and evidence-grounded retrieval in one application. It gives teams a practical way to evaluate agentic video intelligence using their own cameras, terminology, events, and operating requirements.
What the Kit Enables
- Summarize live and recorded video without continuous manual review
- Search footage using events, objects, locations, time periods, or natural-language descriptions
- Create configurable alerts for safety events, restricted access, vehicle activity, and other operational conditions
- Keep sensitive footage, summaries, clips, and metadata under local control
How CTai LABS Helps
CTai LABS created Scene Analyzer Agent as an integrated Edge ai application, not simply an implementation of individual models or NVIDIA frameworks. Our team engineered the workflow across video ingestion, accelerated processing, VLM-based scene understanding, Nemotron-based reasoning, alert generation, local storage, retrieval, and evidence-grounded question answering.
CTai LABS can adapt the kit around the customer’s cameras, data-governance requirements, terminology, objects, locations, alert conditions, and target NVIDIA Jetson™ platform. Model selection, quantization, serving configuration, video processing, memory use, and compute offload can also be optimized for the required number of streams and performance target.
Connect Tech Hardware and Connectivity
The Scene Analyzer Agent Kit is built around Connect Tech Edge ai compute, video-ingestion hardware, storage, and the cables and accessories required for the selected camera configuration.
Anvil-T5 Edge System with NVIDIA® Jetson Thor™
Anvil Embedded System with NVIDIA® Jetson AGX Orin™
| Part Number | Description |
|---|---|
|
ESG625-45Z
OR ESG620-0QX |
Anvil-T5 with Jetson T5000, WiFi 6/BT Module, 1 × 1TB NVMe, and GMSL2 Camera Board
OR Anvil System, Active Cooling, with Jetson AGX Orin Module, WiFi 5/BT Module, 1 × 1TB SSD, and GMSL Camera Board |
| MSG121 | AC/DC PSU, 48V 360W, 2% 6-pin Mini-Fit Jr., for Gauntlet T5000 carrier |
| CBG348 | Quad FAKRA GMSL Cable |
| MSG093 | WiFi/BT Antenna |
Together, these components provide a hardware foundation for local video processing, model inference, alert generation, evidence storage, and natural-language search. CTai LABS configures the system around the required camera inputs, stream count, model size, storage needs, and deployment environment.
Explore the related solution
Learn more about the underlying video-ingestion, VLM, LLM, retrieval, alerting, and local-storage architecture on the Scene Analyzer Agent Solution page.
See the kit in action
Bring CTai LABS your robotics perception challenge, sensor requirements, robot architecture, and target Edge platform.
3D Spatial Intelligence Kit
Turn a Single Camera Stream into 3D Scene Understanding
The 3D Spatial Intelligence Kit transforms a single monocular camera stream into metric depth, colorized point clouds, and structured 3D scene information. On-device spatial reasoning adds object categories, orientation, scale, and spatial relationships to the reconstructed environment.
The kit provides a focused way to evaluate 3D perception without beginning with a stereo rig, LiDAR, or calibrated multi-camera array. Robotics teams can determine how locally generated spatial information supports navigation, obstacle avoidance, manipulation, inspection, SLAM, and planning before developing a larger perception system.
What the Kit Enables
- Generate metric depth and colorized 3D point clouds from a single 2D camera.
- Identify objects, layouts, orientation, free space, and spatial relationships.
- Provide machine-readable scene information to ROS2 robotics applications.
- Evaluate real-time 3D perception in disconnected or bandwidth-constrained environments.
How CTai LABS Helps
CTai LABS integrates camera ingestion, accelerated 3D reconstruction, spatial reasoning, and ROS2 around the customer’s robot, sensors, and deployment environment.
The kit extends the Connect Tech ROS-Ready Launchpad and CTI EdgeAI Stack by connecting GMSL camera ingestion, NVIDIA Jetson compute, TensorRT inference, NVIDIA Isaac™ ROS, point-cloud generation, and spatial-reasoning outputs in one Edge architecture.
Our engineers can adapt the sensor configuration, model selection, ROS2 interfaces, spatial outputs, performance settings, and downstream connections for the customer’s navigation, manipulation, inspection, or planning workflow.
Connect Tech Hardware and Connectivity
The 3D Spatial Intelligence Kit combines Connect Tech Edge ai compute with camera-support hardware, sensors, and connectivity selected for real-time monocular depth estimation and 3D scene reconstruction.
Anvil-T5 Edge System with NVIDIA® Jetson Thor™
| Part Number | Description |
|---|---|
| ESG625-45Z | Anvil-T5 with Jetson T5000, WiFi 6/BT Module, 1 × 1TB NVMe, and GMSL2 Camera Board |
| MSG121 | AC/DC PSU, 48V 360W, 2% 6-pin Mini-Fit Jr., for Gauntlet T5000 carrier |
| CBG348 | Quad FAKRA GMSL Cable |
| MSG093 | WiFi/BT Antenna |
The configured system provides the camera-to-compute foundation for generating depth maps, colorized point clouds, and structured spatial information at the Edge. CTai LABS aligns the hardware, sensor interfaces, accelerated software, and ROS2 outputs with the customer’s robotics platform.
Explore the related solution
Learn more about monocular depth estimation, real-time point-cloud generation, spatial reasoning, and ROS2 connectivity on the 3D Spatial Intelligence Solution page.
See the kit in action
Bring CTai LABS your robotics perception challenge, sensor requirements, robot architecture, and target Edge platform.
Physical ai Evaluation Kit
Connect the Physical World, Digital Twins, and Edge Deployment
The Physical ai Evaluation Kit connects live sensor data, digital twins, foundation models, robotics, and Edge deployment in a working Real2Sim-Sim2Real environment. Real-world data is processed at the Edge and projected into a simulation where teams can evaluate scene understanding, robot behaviour, and model policies.
Rather than treating simulation and physical deployment as separate activities, the kit helps teams examine how information moves between them. Engineering teams can test a Physical ai architecture on real hardware, identify technical gaps, and gather evidence before investing in a larger custom system.
What the Kit Enables
- Stream live multi-camera tracking and spatial data into a digital twin.
- Evaluate scene understanding and robot behaviour under repeatable simulated conditions.
- Compare simulated results with real-world system performance.
- Explore the movement of validated models and policies from simulation to Edge hardware.
How CTai LABS Helps
CTai LABS integrates the hardware, sensor architecture, simulation environment, models, and Edge deployment workflow around the customer’s evaluation goals.
The demonstrated configuration connects the Connect Tech Anvil-T5, NVIDIA Jetson T5000™, NVIDIA DeepStream, NVIDIA Isaac Sim, Omniverse, Scene Analyzer Agent, Cosmos 3 Edge, Nemotron, Isaac GR00T, and NVIDIA DGX Spark in one working pipeline.
CTai LABS can help define the evaluation target, configure the physical and simulated environments, connect live data to the digital twin, establish test criteria, and plan how validated models or policies move back to Edge hardware. The result is evidence for the customer’s next engineering decision, not simply a standalone simulation.
Connect Tech Hardware and Connectivity
The Physical ai Evaluation Kit brings together Connect Tech Edge ai hardware, camera-support components, sensors, and connectivity for capturing and processing the real-world data used throughout the Real2Sim-Sim2Real workflow.
Anvil-T5 Edge System with NVIDIA® Jetson Thor™
| Part Number | Description |
|---|---|
| ESG625-45Z | Anvil-T5 with Jetson T5000, WiFi 6/BT Module, 1 × 1TB NVMe, and GMSL2 Camera Board |
| Varies by distributor | NVIDIA DGX Spark |
| MSG121 | AC/DC PSU, 48V 360W, 2% 6-pin Mini-Fit Jr., for Gauntlet T5000 carrier |
| CBG348 | Quad FAKRA GMSL Cable |
| MSG093 | WiFi/BT Antenna |
These components connect physical data collection and Edge processing to the wider simulation and robotics environment. CTai LABS configures the hardware and sensor architecture around the digital twin, model-evaluation requirements, and intended route back to physical deployment.
Explore the related solution
Learn more about the Real2Sim-Sim2Real architecture, live digital-twin workflow, scene understanding, and autonomous manipulation on the Physical ai Evaluation Kit Solution page.
See the kit in action
Bring CTai LABS your robotics perception challenge, sensor requirements, robot architecture, and target Edge platform.
ScrapGuard™ Kit
Identify Hazardous Cylinders Before They Reach the Shredder
The ScrapGuard™ Kit is a site-configured Edge ai system for identifying sealed pressurized cylinders in heterogeneous scrap streams before they reach a downstream shredder. A three-stage perception process evaluates candidate hazard regions and makes verified results available to the customer’s existing PLC and diverter architecture.
The kit brings together rugged Edge compute, industrial vision, local inference, control connectivity, and field hardware as a defined starting architecture. CTai LABS configures the system around representative site data, conveyor conditions, camera placement, target classes, and the available decision window.
What the Kit Enables
- Evaluate earlier visual identification of sealed pressurized cylinders in scrap streams.
- Process camera data locally without cloud dependency in the operational detection path.
- Connect verified perception results to an existing PLC and diverter architecture.
- Establish a measured site baseline using representative materials and operating conditions .
How CTai LABS Helps
CTai LABS can begin with a defined hazard, sample video, a camera preference, a PLC requirement, an existing model, or a complete installation brief. Our team fills the missing requirements, selects the appropriate kit configuration, adapts the three-stage perception stack to representative data, and defines the shortest credible route to a measured site baseline.
Support can include computer vision, camera and lighting selection, rugged Edge compute, PLC interfaces, threshold tuning, site validation, documentation, and handover. The customer retains its source data, proprietary process knowledge, application, system design, and project IP.
ScrapGuard™ provides a perception result and an engineered connection to the customer’s control architecture. It does not replace the facility’s risk assessment, HAZOP process, certified safety functions, guarding, interlocks, or final operational authorization.
Connect Tech Hardware and Connectivity
The ScrapGuard™ Kit combines the Connect Tech Anvil-T5 Edge system with an industrial GMSL camera, specified cabling and interconnects, mounting and field hardware, industrial power, and a PLC-facing interface.
Anvil-T5 Edge System with NVIDIA® Jetson Thor™
| Part Number | Description |
|---|---|
| ESG625-45Z | Anvil-T5 with Jetson T5000, WiFi 6/BT Module, 1 × 1TB NVMe, and GMSL2 Camera Board |
| Varies by distributor | NVIDIA DGX Spark |
| MSG121 | AC/DC PSU, 48V 360W, 2% 6-pin Mini-Fit Jr., for Gauntlet T5000 carrier |
| CBG348 | Quad FAKRA GMSL Cable |
| MSG093 | WiFi/BT Antenna |
The final configuration is defined around the site’s conveyor geometry, camera position, lighting, line speed, PLC architecture, and available decision window. CTai LABS supplies the approved hardware and software baseline as a coordinated package, reducing on-site assembly while preserving the calibration and validation required for the operating environment.
Explore the related solution
See how CTai LABS developed and tested the three-stage perception system using real customer scrap data on the ScrapGuard™ Recycling Explosion Detection page.
See the kit in action
Bring CTai LABS your robotics perception challenge, sensor requirements, robot architecture, and target Edge platform.
Start with a kit. Build toward your application.
Whether you are evaluating video intelligence, spatial reasoning, a Real2Sim-Sim2Real workflow, or industrial hazard detection, CTai LABS can help turn a focused starting point into a measured engineering plan.