As artificial intelligence finds its way into engineering workflows, Keysight Technologies is taking an approach that keeps engineers at the center of the design process. Rather than using AI to replace established design methodologies, the company is focused on helping engineers work more efficiently by capturing engineering knowledge, automating repetitive tasks, and accelerating computationally intensive workflows while continuing to rely on proven simulation and verification tools.
This strategy is built around three pillars:
- Capture: Preserve engineering knowledge, workflows, and design decisions so they can be reused across projects and teams.
- Orchestrate: Connect design tools, simulations, and workflows to reduce manual effort and improve productivity.
- Accelerate: Speed up design exploration and optimization by reducing the computational bottlenecks associated with complex simulations.
For RF engineers, this approach is designed to build on existing tools rather than replace them. Design environments such as ADS and Nexus already provide the simulation, optimization, verification, and modeling capabilities needed to design RF and microwave systems. The challenge is that much of the engineering knowledge behind those designs remains locked inside projects, while the simulations needed to validate them can be time-consuming and computationally expensive. SOS AI is intended to bridge that gap by helping organizations preserve expertise and apply automation around proven design flows.
Understanding the Simulation Challenge
One of the biggest bottlenecks in RF and microwave design is electromagnetic (EM) simulation. Engineers rely on EM simulations to predict how antennas, filters, packages, interconnects, and matching networks will behave before hardware is manufactured. While these simulations provide highly accurate results, they can also require substantial computing resources. A designer may need to evaluate hundreds or even thousands of design variations during development, but running a full EM simulation for every iteration quickly becomes impractical. As designs become more complex, the ability to explore additional design options is often limited by simulation time rather than engineering creativity.
This is the problem that much of Keysight’s AI strategy is trying to solve. Rather than replacing engineering judgment, the company is focused on helping engineers explore more of the design space, automate repetitive processes, and reuse knowledge that already exists within their organizations.
Understanding Surrogate Models
At the heart of SOS AI is the concept of a surrogate model. A surrogate model is a fast approximation of a detailed EM simulation. To understand why this is useful, consider an antenna, filter, or package design that requires a full EM simulation every time a parameter changes. While these simulations accurately model the underlying physics, they can also take significant time to complete.
To create a surrogate model, engineers first run many high-fidelity EM simulations across a broad range of design conditions. The resulting data is then used to train a model that learns how the original design behaves. Once validated, that model can predict performance much more quickly than the original simulation.
The surrogate-modeling workflow consists of five main steps:
- Run a large number of high-fidelity EM simulations across the design space.
- Collect and organize the resulting simulation data.
- Create and validate a surrogate model based on that data.
- Use the surrogate model to rapidly explore design alternatives and optimize performance.
- Replace the surrogate with the original EM simulation for final verification.
In practice, the surrogate acts as a fast stand-in during design exploration and optimization. Once engineers identify a promising solution, they return to the original physics-based simulation for final validation. This allows them to investigate significantly more design alternatives without sacrificing confidence in the final result.
Capture: Turning Engineering Expertise into Reusable Workflows
Valuable engineering expertise often exists in design files, simulation setups, specifications, whiteboard sketches, scripts, and the experience of individual engineers. While these assets are critical to successful product development, they are often difficult to reuse across projects or engineering teams. The Capture pillar focuses on preserving this expertise and making it reusable. During its demonstrations, Keysight described capabilities that can capture design workflows, record engineering actions, and transform existing workflows into reusable Python-based processes. The company also highlighted methods for capturing design intent from whiteboards and workflow diagrams, allowing engineering processes to be preserved alongside the final design itself.
Rather than simply storing a completed design, the goal is to preserve how that design was developed. For example, simulation setups, optimization strategies, design decisions, and workflow sequences can be captured and converted into structured processes that other engineers can reuse later.
This is particularly important because Python has become a common language for automation and AI workflows. By converting engineering processes into reusable Python-based workflows, Keysight aims to make decades of engineering expertise accessible to future automation systems while keeping the original engineering methodology intact.
Orchestrate: Connecting Tools, Data, and Design Flows
Once knowledge has been captured, it must be connected to the broader engineering environment. The Orchestrate pillar focuses on linking design tools, simulation engines, engineering data, and automation frameworks into a unified workflow. Instead of requiring engineers to manually launch simulations, export data, train models, validate results, and manage files across multiple tools, SOS AI can coordinate those activities through automated pipelines.
Within the surrogate-modeling workflow, orchestration manages the entire process of generating data, training models, validating accuracy, and deploying approved models. Engineers define the objectives and constraints, while the platform handles much of the workflow execution behind the scenes.
Keysight has emphasized that this is intended to work alongside existing design environments rather than replacing them. The company’s broader strategy is to connect automation and AI capabilities directly to proven engineering toolchains so organizations can continue using their established design methodologies.
Accelerate: Scaling Design Exploration with Surrogate Models
The Accelerate pillar is where SOS AI delivers its most visible benefit. Traditional EM simulations provide excellent accuracy but can limit the number of design alternatives engineers are able to evaluate during a project. Surrogate models help remove that limitation by providing a much faster approximation of the original simulation.
Creating a surrogate begins with generating simulation data across a broad design space. SOS AI manages that process, collects the resulting data, and uses it to create a surrogate model that reproduces the behavior of the original design. Once validated, the model can be used to perform rapid optimization, investigate additional design variations, and explore a wider range of operating conditions.
These surrogate models can then be stored as reusable assets and made available to other designers through a shared library. Instead of repeatedly rebuilding models for similar design challenges, organizations can reuse validated assets and build upon previous simulation efforts. The result is faster iteration, broader design-space exploration, and more opportunities to identify optimal solutions before moving back to the original high-fidelity simulation for final verification.
Beyond Data Management
While SOS originally functioned as a data-management system, SOS AI expands its role considerably. Beyond storing engineering information, it provides tools for managing experiments, automating simulation pipelines, creating surrogate models, validating results, and governing engineering assets throughout their lifecycle. The platform maintains version-controlled datasets, models, and design assets connected through a digital thread. This allows engineers to trace a surrogate model back to the exact design version, simulation data, and training process used to create it. Such traceability helps build confidence in model quality while making validated assets easier to reuse across teams and projects.
Through SOS AI, Keysight’s broader AI strategy becomes clear. Rather than positioning AI as a replacement for engineers, the company is applying it to specific bottlenecks in the design process. Engineers remain responsible for defining problems, making design decisions, and validating results, while AI and automation help capture expertise, orchestrate workflows, and accelerate simulation-driven design.
Keysight Technologies recently highlighted this approach and demonstrated the SOS AI surrogate-modeling workflow during European Microwave Week 2026.
Click here to learn more about planning your AI design journey in RF EDA.