NEPI Platform Overview
Executive Summary
Building a smart system is hard. Not because the application is complex, but because teams spend months on infrastructure before they can write a single line of application code. NEPI eliminates that problem. It is an open-source edge AI platform that handles sensor integration, AI model deployment, event-driven automation, and data collection out of the box. Teams start building what actually matters from day one.
The Challenge
Smart systems have never been more in demand. Across robotics, autonomous vehicles, drones, subsea platforms, and industrial automation, the ability to collect data, run AI on-device, and act on the results in real time is becoming table stakes.
But building that capability from scratch is a different story.
Every smart system requires a foundation of low-level software that most teams have no interest in building: hardware drivers, sensor management, AI runtime environments, data pipelines, and control interfaces. These components are not differentiating. They do not make the product better. But they consume months of engineering time and some of the most specialized talent available.
As a result, teams face a common pattern:
- Six to twelve months building infrastructure before the actual application work begins
- Hardware changes that force rewrites of custom drivers and integration code
- Brittle data pipelines that work in the lab but fail in the field
- Difficulty hiring for the deep, specialized roles required to maintain this infrastructure
- Limited ability to reuse any of this work across projects
Meanwhile, the window for getting to market is shrinking. The teams that ship fastest win. And the teams shipping fastest are not the ones with the best infrastructure engineers. They are the ones that did not have to build infrastructure at all.
The Solution: NEPI
NEPI (Numurus Edge Platform Interface) is an open-source software platform that gives engineering teams the infrastructure layer that most smart systems need out of the box.
Think of it like the operating system layer for smart systems. Before Windows, building a software application meant understanding every piece of hardware and writing code to make it function. NEPI does the same thing for smart systems: it removes the infrastructure barrier so teams can build what they actually shipped to build.
The 90/10 Rule
NEPI handles approximately 90% of what most smart systems need out of the box: hardware drivers, AI model management, event-driven automation, data collection, and a browser-based UI. Teams customize the remaining 10% for their specific application. Their own AI models, automation logic, and application layer. That 10% is where the product lives. Everything else is handled.
NEPI is built on ROS and ROS 2, the most widely used frameworks in robotics and autonomous systems. Teams already in the ROS ecosystem can adopt NEPI without leaving behind existing code or knowledge. The full codebase is on GitHub at github.com/nepi-engine.
NEPI runs fully at the edge. No cloud dependency. No connectivity requirement. This is critical for the environments where smart systems are actually deployed: underwater, in the field, on autonomous vessels, in connectivity-denied locations.
What NEPI Does
NEPI is made up of three tightly integrated components that together create a complete development and deployment environment for edge AI systems.
Hardware Abstraction Drivers
NEPI provides a growing library of hardware drivers that abstract each device's native interface to a NEPI standard SDK. Applications connect to NEPI, not to specific hardware. That means swapping out a camera, sonar, or GPS unit does not require rewriting any application code.
Supported hardware categories include:
- Cameras (RGB, depth, stereo, infrared)
- Sonar and lidar
- Pan-tilt systems
- GPS and navigation devices
- Lighting and strobe systems
- Robotic control systems
AI Model Management
NEPI includes a drag-and-drop AI model deployment system that abstracts each model's native framework interface to a standard SDK. Teams can deploy, swap, and update AI models without touching the rest of the system. Live data streams connect directly to custom or off-the-shelf AI models that feed automation decision-making in real time.
Event-Driven Automation and Data Collection
NEPI provides a low-code automation system that lets teams define triggers, conditions, and actions without writing infrastructure code from scratch. The event-driven data collection system ties sensor states, AI detections, and time-based triggers together into structured, high-quality data pipelines that work in the field, not just in the lab.
Browser-Based User Interface
NEPI's Resident User Interface (RUI) is a customizable browser-based interface hosted directly on the NEPI device. No custom front-end development required. Teams can configure hardware drivers, validate sensor connections, manage AI models, and monitor system behavior from any browser on the network.
Applications and Use Cases
NEPI is hardware-agnostic and industry-agnostic. The same platform that runs maritime threat detection on an autonomous surface vessel also powers inspection automation on an underwater ROV and provides hands-on AI tooling for student robotics programs.
Smart Sensing
Smart sensing solutions create actionable information, not raw data, in real time at the point of collection. NEPI connects sensors to AI models and automation logic so platforms can detect, classify, and respond without a human in the loop. This reduces operator burden, improves response time, and enables platforms to operate in environments where continuous human oversight is not practical.
Smart Monitoring
Monitoring solutions built on NEPI can detect specific conditions and trigger responses automatically: logging events, sending alerts, tracking objects, activating systems, or routing information to remote endpoints. NEPI's event-driven architecture means monitoring behaviors can be defined and updated without rebuilding the underlying platform.
Smart Inspection
Inspection is expensive when done manually. The cost of a missed data point on the first pass can be significant. NEPI-enabled inspection systems increase the quality of critical data collection while reducing the volume of non-critical data, whether the inspection is performed by a human operator or an autonomous robotic system.
STEM and Robotics Education
One of the most significant barriers in robotics education is the gap between the skills students need to learn and the infrastructure complexity required to get to those skills. NEPI eliminates that gap. Student teams can connect sensors, deploy AI models, and build automation workflows without needing to understand low-level hardware interfacing or ROS middleware from the ground up.
Industry Applications
NEPI has been deployed across a range of demanding environments and industries. The platform handles the infrastructure layer the same way regardless of application, which means teams in different sectors can use and benefit from the same core tools.
| Sector | Application |
|---|---|
| Robotics and Autonomous Systems | Sensor integration, AI model deployment, and automation for ground, aerial, surface, and subsea platforms. Reduce development time from months to days. |
| Defense | Autonomous sensing and decision-making in connectivity-denied environments. Rapid deployment of AI-enabled inspection and monitoring capabilities on field platforms. |
| Industrial | Automated monitoring and inspection operations using edge AI. Real-time actions without a human-in-the-loop. Short project timelines and open-source flexibility. |
| Science and Research | Automated, real-time field data collection workflows. Simplified sensor setup and data processing in remote locations, so researchers can focus on discovery. |
| Education | Hands-on AI and robotics project tooling for student teams. Leapfrog infrastructure complexity to focus on learning the skills that matter. |
The NEPI Container
NEPI is now available as a Docker container that runs on any standard Linux machine, including a standard laptop or desktop. This removes the last remaining barrier to getting started.
Previously, running NEPI required dedicated edge compute hardware. The NEPI Container eliminates that requirement. Any team with a Linux machine and 60GB of free space can install and run the full NEPI platform immediately, with the same capabilities available on hardware deployments.
What the NEPI Container means in practice
Student teams can run NEPI on a laptop at the mission station. Researchers can prototype AI and sensor workflows before hardware arrives. Startups can evaluate the full platform before committing to a hardware configuration. The container is the same platform. It just runs on the hardware you already own.
Key facts:
- Runs on any standard Linux machine
- No dedicated edge hardware required
- Full NEPI capabilities, containerized
- 60GB free disk space required
- Installation instructions at github.com/nepi-engine/nepi_setup
Software Architecture
NEPI installs on top of the base operating system provided by edge-compute board and chip manufacturers. It can be installed directly on a device's OS or run as a Docker container within the device's OS. The direct installation option provides additional clock, network, security, and software management features. The container option provides a faster setup path and easier portability across hardware platforms.
Primary NEPI command-line and programmatic interfacing is provided by a collection of ROS communication primitives and shared interface classes. NEPI supports multiple API interfaces, with full documentation at nepi.com/documentation.
The software system includes three main component types:
- System and management nodes: handle network, drivers, data, software, AI, and application management
- Resident User Interface (RUI): browser-based control and monitoring interface hosted on the device
- Shared user storage: network-accessible storage drive for data access and application development
Numurus Products and Services
Numurus builds and supports the NEPI platform and offers a range of products and services to help teams move from evaluation to deployment.
| Product / Service | Description |
|---|---|
| NEPI Software (Open Source) | Full platform codebase available on GitHub. Free to install, modify, and deploy. Built on ROS and ROS 2. |
| NEPI Container | Docker-based NEPI image for standard Linux machines. No dedicated hardware required. |
| Turn-Key NEPI Images | Pre-built NEPI software images with optimized Ubuntu OS, hardware drivers, AI frameworks, and applications. Ready to deploy on supported hardware. |
| NEPI-Ready Hardware (S2X) | Off-the-shelf and custom edge-compute hardware stacks pre-installed with NEPI software and drivers. Available in multiple configurations for different deployment environments. |
| Professional Services | AI, robotics, and smart sensing engineers available for project support, prototyping, and custom solution development. |
Developer Resources
NEPI is built for teams that want to move fast. All core resources are publicly available.
GitHub
Full NEPI source code, installation instructions, and SDK at github.com/nepi-engine. Open source, fork-friendly, and actively maintained.
Documentation and Tutorials
Comprehensive documentation, integration guides, API references, and step-by-step tutorials at nepi.com/documentation.
Video Library
Feature walkthroughs, use case demonstrations, and development tutorials on the NEPI YouTube channel at youtube.com/@Numurus-NEPI.
Community
The NEPI Community Discord and forum at community.nepi.com provides a space for users and developers to get answers, share work, and discuss upcoming features. Permanent Discord invite: discord.gg/7WXgVUgXmX
Customer Success: OceanAero
The combination of Numurus' NEPI smart system software, off-the-shelf compute hardware, and its responsive engineering support team was a big factor in the success of this project. It saved our internal team from a lot of development work we'd otherwise have to do ourselves.
Kevin Decker, CEO | OceanAero
OceanAero, maker of the TRITON autonomous underwater and surface vehicle (AUSV), needed to deploy AI-enabled inspection and threat detection capabilities on its platforms in support of the Defense Innovation Unit's Unmanned Systems for Maritime Domain Awareness program.
The goal: automate 360-degree maritime threat detection using on-board cameras and AI models, with detected threat information transmitted wirelessly to remote operations centers. Within six months, their team had:
- Interfaced five directional cameras with on-board AI models
- Fielded, tested, and demonstrated automated maritime domain awareness on the TRITON AUSV
- Reduced dependency on human-in-the-loop operations for threat detection
- Delivered actionable information to local and remote endpoints in real time
More case studies at nepi.com/case-studies, including VideoRay, WESMAR Smart Sonar, UWT Ferry Project, and the LLC Drone Capstone program.