In our ongoing series on agentic workflows with Dyad, we have highlighted the use of Dyad’s AI agent to create, refine, and analyze system models for model-based systems engineering. In the first episode, we used Dyad’s AI agent to develop and calibrate a multi-domain friction brake model, demonstrating how agentic workflows can transform model development into a faster, more iterative, and insight-driven process.
In the latest videos, we tackle a new workflow- namely developing a coffee cup thermal model from minimal input data. The agent begins with nothing more than a schematic and a few simulation plots originally created by Dr. Clément Coïc. From these limited artifacts, the agent interprets the image, identifies the relevant physics and components, and generates the Dyad model to simulate the transient cooling of the coffee. We even calibrate the model using the agent to match existing results.
Watch the video to see how Dyad’s agent creates and simulates the coffee cup thermal model in real time.
In the followup video, we continue using Dyad’s AI agent to restructure the thermal model and add physics.
Of course, we aren’t typically building models from images of models. But this image was simply a proxy for the kinds of graphical representations that are commonly used as the starting points for modeling activities - engineering drawings, schematics, sketches in a notebook, PowerPoint drawings, etc. In the video, we even explore interactions with the agent on a more abstract image.
What makes this capability significant is the sophistication of the workflow. Dyad’s AI agent bridges the gap between conceptual understanding and computational modeling, translating images, symbols, and contextual cues into executable, production-ready code. It doesn’t just recognize components; it understands the underlying physics, relationships, and governing equations that define the system’s behavior and can provide relevant engineering insight based on this information starting from model creation and continuing through the simulation of the model.
This capability marks a significant shift in how models are built and validated. Agentic workflows offer potential gains in efficiency with less manual work, iterations, and specialized expertise (but let’s be honest, the engineers still need to know what they are doing!!). Dyad’s agentic approach demonstrates how AI can compress that workflow, automating interpretation, generation, and testing, allowing engineers to move more seamlessly from idea to simulation to engineering analyses.
Beyond the immediate application, this example portends to a broader shift in engineering workflows. As AI becomes more capable of understanding design intent and physical reasoning, tools like Dyad will redefine what it means to model, simulate, and iterate. Instead of focusing on the coding and creation of the model, engineers can focus on higher level engineering workflows such as architectural exploration, validation, and optimization while AI handles the translation from prompts to executable code.
The result is not just efficiency. It’s a new kind of engineering collaboration between human insight and machine intelligence.
