JuliaLang Dyad v2.0.0 Enables AI to Propose & Test Engineering Experiments
Summarize this article with:
JuliaHub announced the release of Dyad v2.0.0, bringing together agentic AI and simulation to create an environment where models can act as interactive collaborators. This new release enables models to propose formulations, generate experiments, test hypotheses, and autonomously refine results, effectively closing the loop between reasoning and simulation. By facilitating this process, Dyad accelerates engineering workflows and supports verified model development, representing a major step toward AI-assisted modeling. JuliaHub is also partnering with Synopsys to integrate Dyad with Ansys TwinAI, further enhancing digital twin technology. Dyad v2.0.0 and Agentic AI Capabilities Dyad v2.0.0 introduces powerful agentic AI capabilities, bringing together agentic AI and simulation in a seamless environment. This allows models to act as interactive collaborators, proposing formulations, generating experiments, and autonomously refining results. By closing the loop between reasoning and simulation, Dyad accelerates engineering workflows and supports verified model development, moving away from manual trial-and-error processes. Demonstrations showcase Dyad’s agentic AI in action, including building a thermal model of a turkey to explore cooking strategies and creating a thermal model of a coffee cup directly from a schematic and sample plots. These examples highlight the ability of the AI agent to interpret visuals, extract relevant physics, generate executable code, and calibrate models – all in real-time, saving significant engineering time. JuliaHub will present Dyad’s Agentic AI capabilities at the AIAA SciTech Forum, taking place January 12-16, 2026. This event, attracting over 6,100 attendees, will provide a platform to showcase the technology and explore its applications within the aerospace R&D community. Interested parties are encouraged to connect with JuliaHub representatives during the forum. Dyad Integration with Ansys TwinAI and SciML Dyad v2.0.0 integrates agentic AI with simulation, enabling models to act as interactive collaborators. This new release allows models to propose formulations, generate experiments, and autonomously refine results, accelerating engineering workflows and supporting verified model development. A key component of this integration is the introduction of Dyad’s graphical interface, designed for both exploratory modeling and scalable engineering applications. A significant partnership brings Dyad into Ansys TwinAI, combining JuliaHub’s AI-driven, physics-informed simulation expertise with Synopsys’ digital twin technology. Dr. Prith Banerjee of Synopsys highlights that this integration empowers engineers to build evolving digital twins, bridging the gap between simulation and real-world data. This collaboration aims to accelerate innovation and enhance accuracy in hardware design and system optimization. Dyad’s capabilities extend to model discovery through tools like Universal Differential Equations (UDEs). This allows uncovering missing physics within models, and the insertion of neural networks to extract symbolic representations of learned dynamics. Demonstrations include building a thermal model of a coffee cup from a schematic, interpreting visuals and generating executable code in real-time, showcasing faster iteration and reduced manual coding. Dyad Applications: Modeling and Simulations Dyad v2.0.0 introduces agentic AI capabilities, bringing together AI and simulation in a seamless environment. This allows models to act as interactive collaborators, proposing formulations, generating experiments, and autonomously refining results. By closing the loop between reasoning and simulation, Dyad accelerates engineering workflows and supports verified model development, moving beyond manual trial-and-error processes. This advancement promises a major step toward AI-assisted modeling. Dyad’s capabilities extend to building and analyzing complex systems, demonstrated by a turkey cooking model and a coffee cup thermal model. The AI agent can build thermal models from sketches or diagrams, generating executable code and calibrating models in real-time. This saves engineers time and unlocks faster iteration, while also showcasing the potential to optimize real-world systems, from holiday dinners to aerospace applications. Furthermore, Dyad is being integrated with Synopsys’ Ansys TwinAI, leveraging SciML technology to build evolving digital twins. Early deployments utilizing physics-based models have reported over 90% prediction accuracy in areas like predictive maintenance and emissions reduction for water operations. This integration aims to bridge the gap between simulation and reality, enhancing hardware design and system optimization. We help enterprises build deployable and scalable solutions leveraging SciML to create highly accurate and trustworthy Digital Twins.JuliaHub Julia Programming Language Updates and Releases JuliaHub announced the release of Dyad v2.0.0, integrating agentic AI and simulation for accelerated engineering workflows. This new version allows models to act as interactive collaborators, proposing formulations, generating experiments, and autonomously refining results. A key feature is its graphical interface, designed for both exploratory modeling and scalable engineering. Engineers can now leverage Dyad’s AI to build models from sketches or diagrams, saving time and unlocking faster iteration, as demonstrated with a thermal model of a coffee cup. The Dyad platform is also being integrated with Synopsys’ Ansys TwinAI, combining JuliaHub’s AI-driven physics-informed simulation with digital twin technology. This partnership aims to accelerate innovation and enhance the accuracy of hardware design and system optimization. Furthermore, JuliaHub is showcasing Dyad’s agentic AI capabilities at the AIAA SciTech Forum, attracting over 6,100 attendees, and providing a venue for connection and demonstration of the technology. Recent developments include the release of Julia v1.12.2, offering bug fixes, documentation cleanup, and performance improvements, and the launch of the Julia Security Working Group (JLSEC) to strengthen Julia’s security tooling.
The Julia Ecosystem Benchmarks Explorer provides interactive visual reports of package loading and precompilation times across Julia versions, aiding performance analysis. Several webinars and newsletters are also available, including topics on causal modeling and building Julia binaries. JuliaHub Events: Webinars, Conferences, and Livestreams toJuliaHub hosts a variety of events including webinars, livestreaming sessions, and conference presentations. Weekly livestreams feature Dr. Chris Rackauckas building real-world models with Dyad’s agentic AI every Friday at 10am EST on JuliaHub YouTube and @ChrisRackauckas Twitch. Additionally, JuliaHub provides free one-hour webinars led by staff and experts; recent topics include causal vs. acausal modeling and building Julia binaries. Nearly 100 past webinars are available online for on-demand viewing. The company actively participates in larger industry events, such as the AIAA SciTech Forum in Orlando, Florida (January 12-16) where they will present Dyad’s Agentic AI capabilities. JuliaCon 2026 is already scheduled for August 10-15, 2026, at Johannes Gutenberg University in Mainz, Germany. JuliaHub also offers access to recordings of talks from JuliaCon 2025, covering topics like standalone executables and AI’s impact on Julia development. JuliaHub leverages these events to showcase Dyad’s capabilities, including demonstrations of agentic AI modeling—like simulating a turkey’s cooking process or building a thermal model of a coffee cup. They also highlight SciML technology’s potential in areas like predictive maintenance and emissions reduction, reporting over 90% prediction accuracy in early deployments, and offer resources like the Julia Ecosystem Benchmarks Explorer for performance analysis.
Scientific Machine Learning and Modeling Techniques Dyad v2.0.0 introduces agentic AI capabilities, seamlessly integrating AI and simulation. This allows models to act as interactive collaborators, proposing formulations, generating experiments, and autonomously refining results. Demonstrations include building a thermal model of a turkey and a coffee cup from sketches, showcasing rapid model generation and iteration. This accelerates engineering workflows and supports verified model development, moving beyond manual trial-and-error.
Scientific Machine Learning (SciML) is highlighted as key to re-engineering maintenance and reducing emissions in water operations. Deployments report over 90% prediction accuracy in anticipating pump failures and optimizing energy use, signaling a significant shift for asset health management. Furthermore, SciML, utilizing universal differential equations, can construct high-fidelity ODE surrogate models that accurately approximate the outputs of agent-based epidemic simulations. JuliaHub is advancing SciML through partnerships and tools like Dyad. Integration with Synopsys Ansys TwinAI brings together AI-driven physics-informed simulation with digital twin technology. Additionally, Dyad Model Discovery leverages Universal Differential Equations to uncover missing physics within models. These advancements are being presented at events like AIAA SciTech Forum, demonstrating innovation and collaboration in the field. Dyad combines physics-based modeling with scientific machine learning (SciML) for mission-critical engineering. Source: https://juliahub.com/blog/december-2025-newsletter Tags: