On June 9, 2025, at Intel Corporation’s headquarters in Santa Clara, California, Graiphic had the unique opportunity to showcase SOTA, our groundbreaking deep learning and computation framework, at the ONNX Annual Meetup 2025.
Youssef Menjour, CTO of Graiphic, was invited by the ONNX Technical Steering Committee to present our innovative approach that blends visual programming, deep learning, and general-purpose graph computing through LabVIEW, ONNX, and ONNX Runtime.
Watch Youssef’s full presentation from the event here:
Why ONNX & ONNX Runtime Matter in Today’s Technological Landscape
The Open Neural Network Exchange (ONNX) has become a cornerstone of modern AI infrastructure. It provides a universal and open format for representing machine learning models, facilitating seamless portability across various hardware platforms and execution environments.
ONNX Runtime, a powerful, performance-oriented engine, enables optimized inference and training across CPUs, GPUs, and specialized hardware. By abstracting the complexity associated with heterogeneous computing, ONNX Runtime has become essential for deploying models efficiently at scale.
Graiphic’s Vision: Unifying AI and Graph Computing with SOTA
Unlike traditional AI development workflows fragmented by scripts, conversions, and wrappers, SOTA provides engineers, researchers, and academic teams a unified and visual development environment built entirely upon ONNX and ONNX Runtime.
A Revolutionary Approach:
-
No Python, Docker, or Jupyter required:
Pure graphical data-flow programming through LabVIEW, known for its deterministic execution and intuitive debugging. -
Interactive Visual Programming:
Users can visually construct, edit, and orchestrate ONNX computation graphs directly within LabVIEW, enhancing productivity and clarity. -
General-Purpose Graph Computing:
SOTA extends beyond AI, supporting general computational graphs for control systems, measurement applications, and high-performance industrial scenarios.
Discover our Deep Learning Orchestration with ONNX/ONNX Runtime:
Technological Firsts Presented at the ONNX Meetup
During the presentation, Youssef introduced groundbreaking achievements made possible by SOTA:
-
First complete training of a YOLOv11 segmentation model fully within ONNX Runtime.
-
Implementation of a complete Deep Deterministic Policy Gradient (DDPG) reinforcement learning agent, orchestrated visually in LabVIEW through ONNX graphs.
SOTA empowers developers to execute complex AI workflows entirely in a graphical environment, eliminating dependency on Python-based libraries or external frameworks.
General Graph Computation with ONNX Runtime and LabVIEW:
By extending ONNX beyond traditional AI inference and training, SOTA introduces a revolutionary approach to computational graph execution in LabVIEW—applicable across various industries and applications:
Watch how SOTA orchestrates general-purpose computation graphs:
Three Domains Where SOTA Makes a Difference:
- Industrial Automation:
Direct integration of ONNX models into real-time automation frameworks (State Machines, QMH, Actor Frameworks), providing deterministic and high-performance execution for embedded applications. - Academic Education:
An intuitive, visual programming interface combined with high-level APIs drastically reduces barriers to entry for AI education, facilitating faster adoption and deeper understanding. - Research & Development:
Fine-grained control of computational graphs enables researchers to prototype complex custom architectures rapidly, opening new frontiers in experimental AI design.
Proud Member of the ONNX Community
We are incredibly honored to be part of the ONNX community, working alongside technology leaders like Intel, Microsoft, AMD, NVIDIA, National Instruments, and the Linux Foundation. Our commitment is to enhance the accessibility, modularity, and practicality of ONNX-based computing for all developers, researchers, and engineers through LabVIEW.
Special thanks to the ONNX Steering Committee—Alexandre Eichenberger, Ganesan Ramalingam, Saurabh Tangri, Mayank Kaushik, and Andreas Fehlner—for their support and trust.
As recently announced at the ONNX Annual Meetup 2025, ONNX 2.0 will further advance the generative AI domain, addressing complex deployment scenarios across cloud and edge computing, a vision we deeply share and support.
We look forward to the continued collaboration, innovation, and growth of this outstanding community.
