Sensor Fusion × Physical AI: Enabling Data Fusion and Decision Making at the Edge
In industrial settings, equipment data comes from sensors with different protocols, formats, and frequencies. NeoFlow aggregates, cleans, and transforms this heterogeneous data in real time at the edge, and cross-compares it with the inference results of the Edge AI model to form a cross-modal intelligent decision-making process—completing the closed loop from perception to action on-site without uploading to the cloud or waiting for a response.
Through NeoFlow's innovative and secure process orchestration and synchronization technology, enterprises can uniformly manage the lifecycle of OT data and AI models, truly bridging the information gap between OT and IT, and realizing the implementation of Physical AI on the factory floor.
NeoFlow, a user-friendly workflow orchestration tool, allows for customized enterprise-specific Physical AI.
No-code graphical interface
A node-based No-Code logic engine built on a containerized architecture allows OT engineers to connect data flows by dragging and dropping nodes, eliminating complex development processes and terminal commands. Each node can be configured with processing logic such as conditional judgments, data merging, type conversion, and time window calculations. The system provides real-time prompts for data format mismatches, significantly reducing the risk of configuration errors. For advanced logic needs, custom applications (Custom Apps) can also be deployed, balancing ease of use with flexibility.
Scalable open architecture
Users can develop custom containers (custom apps) using the SDK (supporting Python and Go) to create applications tailored to their specific needs. The platform supports custom nodes, allowing developers to package their own programs into containerized applications and deploy them directly to the edge.
100% Containerization Support
Users can deploy different inference environments at the edge as needed, such as NVIDIA TensorRT, Intel OpenVINO, ONNX Runtime, PyTorch, TensorFlow, or 3rd party AI inference environments, to establish a management mechanism for the integration of edge inference environments and AI model iterations, and quickly deploy AI applications.
Flexible cross-plant deployment with unified scheduling
A single NeoFlow workflow can simultaneously bind dozens to hundreds of industrial gateways. Thanks to its parameterized design, each device uses its own certificate, username, password, and connection information, so engineers do not need to reconfigure the system for each facility and do not have to worry about human errors during the setup process.
When a successful process can be validated on the first production line and only requires parameter adjustments to be quickly replicated across the second and third factories, your industrial AI deployment can move from single-point validation to group-wide implementation.
Use Case
Multinational equipment manufacturers combine AI agents to improve service efficiency
The client is a multinational equipment manufacturer with products distributed throughout the country. In addition to selling products, the company also has maintenance contracts with the client and provides 24/7 service. However, when dispatching maintenance work, it is impossible to obtain real-time information about the equipment's condition on-site. Personnel must go to the site to assess the situation before a repair plan can be determined, resulting in many unnecessary work assignments and increased operating costs.
