Exploring Data & AI Trends in Modeling and Simulation – NAFEMS World Congress 2025 Recap

The NAFEMS World Congress 2025, the largest international conference dedicated to the field of modeling and simulation and digital engineering, brought together users, vendors, consultants, and researchers to explore new technologies, pressing challenges and innovative methods. This year, the event was held in Salzburg, Austria, where vibrant discussions occurred, covering topics ranging from AI for simulation, specialized hardware optimization, and integrating complex software stacks.

Rescale, a proud Gold Sponsor of the event, played a significant role in shaping these discussions. A key highlight was the presentation on”Evolving the Modern Simulation Experience: Orchestrating Compute, Data, and AI for Accelerated Innovation,” delivered by Jacob Surber, VP of Product at Rescale. He was joined by a distinguished panel of industry innovators: George Oates, Thermal Lead Engineer at Rimac Technology, Matthew Chung, Principal Engineer – CAE at Subsea7, and Novid Beheshti, Engineering Team Lead – CAE, CFD at Intelligent Energy. You can watch the panel recording to catch all the insights shared during the panel.

Industry Shifts and Trends in Modeling and Simulation

The combined perspectives from Rescale customers, booth attendees, and presentations across various industries highlighted the challenges, trends, and opportunities CAE engineers are facing today. The landscape of modeling and simulation is undergoing a profound transformation, moving beyond traditional compute-centric workflows to embrace more holistic approaches that integrate data management and AI. While high-performance computing (HPC) remains foundational, the focus has significantly expanded to include GPU-centric workloads and the strategic utilization of simulation data.

One of the most prominent shifts is the increasing complexity of models and designs, demanding increasing multi-disciplinary analysis, computing scale, and flexibility to develop custom AI capabilities. This has led to a greater reliance on cloud computing for its elasticity and ability to provide on-demand access to vast computational resources. Traditional workflows, where engineers perform siloed and often manual handoffs, are being replaced by connected data pipelines that foster collaboration between multiple disciplines. This evolution sees AI moving beyond simple chatbot functionalities, embedding itself into the core development loop of engineering workflows, often leveraging intelligent agents to take actions within these processes. The horizon also shows promise for multi-modality in these models, such as advanced image processing capabilities.

However, common concerns surrounding the security of these sophisticated models and the data being transmitted echo the early conversations about cloud security, presenting a key hurdle for broader adoption that requires clear messaging around data integrity and protection. The strong push for GPU acceleration is evident, driven by the continual desire for faster solution results for simulation workloads compared to running them on traditional x86 CPU architectures. However, its adoption is sometimes constrained by current capacity due to massive AI workload demand and the patchwork GPU coverage across various physics solvers.

Rescale Presents a Holistic Framework for Modern Simulation

Rescale’s VP of Product Jacob Surber presented to a packed room, laying out Rescale’s vision for a modernized simulation workflow built upon three interconnected pillars: Compute, Data, and AI. He emphasized that this holistic framework is designed to empower simulation engineers to overcome current challenges and accelerate innovation.

Compute Orchestration

Compute orchestration focuses on providing elastic cloud HPC to power innovation and maximize efficiency. This involves integrated planning to unify simulation work and manage resources, enabling instant, scalable deployment of a wide range of applications on-demand. The system optimizes workloads by leveraging the latest architectures and AI-driven recommendations for performance and cost-effectiveness. It also offers real-time visualization and monitoring for pre- and post-processing, saving time and cost. While the desire and interest in GPUs for speed-up are high, the current landscape often presents challenges due to capacity constraints and the patchwork GPU compatibility across different physics solvers, which can restrict widespread adoption. The system aims to address this by providing flexible access to diverse hardware. For high-performance storage and collaboration, a shared file system acts as a central repository, supporting demanding AI workflows and facilitating real-time teamwork, data retention, and cost control through tiered storage.

Data Management

The data portion of the presentation emphasized transforming simulation data and processes into automated workflows and insights. A significant observation at the Congress was the consensus that “Data is the problem” in modern simulation, with much of the pain stemming from over a decade of unclean, unstructured data generation. Attendees expressed strong interest in solutions that allow them to bring and leverage their existing data, highlighting a growing demand for data and AI layers that can operate somewhat independently of the core compute offering. This is achieved through a unified data layer that connects to various data platforms, such as SPDM (Simulation Process and Data Management) and PLM (Product Lifecycle Management) systems, to pull and enrich simulation data, enabling comprehensive analysis and report generation and ensuring data integrity. The platform also facilitates automated tasks by reducing manual steps and leveraging engineering process automation to extract key insights, avoiding errors in data entry and hand-off. This framework allows for deploying standardized workflows across an organization and extracting rich metadata crucial for AI capabilities. Additionally, the platform supports AI-enhanced discovery and analysis by automating report write-ups, uncovering hidden insights in job results, and providing recommendations for performance improvements, which is a core function of the Rescale Assistant.

AI-Assisted Modeling and Simulation

The AI pillar focuses on building AI-enhanced simulation on existing methods, continuously improved. This includes a focus on AI for engineering and simulation to enhance overall engineering processes and deliver significant time savings, moving towards truly AI-generated results. Engineers deploying custom models should also focus on building high quality training datasets that feed into AI-based models, utilizing pre-built AI surrogate models, and enabling the development of in-house AI surrogate models. This involves model training and management, leveraging existing simulation data to build AI models, defining tags for data organization, and ensuring traceability for validating AI models. The Rescale platform provides access to the latest GPUs for training and tools for versioning and tracking models, and supports generative capabilities. For model deployment and inference, a dedicated product is used to publish models for inference by collaborators and non-expert users, building a model catalog, and managing model access and versions. For inference, the system integrates with visualization tools to accelerate real-time simulations. Ultimately, the goal is to establish continuous iteration, a workflow where simulation outputs feed AI model training, which in turn informs and refines subsequent simulations, creating a powerful iterative loop for product development.

Live Demonstration: AI-Assisted Data Analysis

A significant highlight of Jacob Surber’s presentation was the live demonstration of a new Rescale Assistant. This beta feature leverages fine-tuned AI large language models (LLMs) to interact with simulation data integrated into the experience of reviewing simulation jobs and result files. The demo showcased Rescale Assistant integrated directly into the simulation interface allowing users to pose natural language queries (e.g., “Show me the relationship between temperature and pressure drops in this aerodynamics study”). The Assistant processed the query, analyzed the underlying simulation data, and generated a real-time chart visualizing the relationship, along with textual analysis and actionable suggestions for further exploration. It also demonstrated multilingual capabilities, providing outputs in different languages. This feature dramatically reduces the time engineers spend extracting insights, allowing them to simply ask questions and let the AI interpret, visualize, and suggest next steps, embodying concepts of abstraction, low code, and no code.

To learn more about a Rescale’s roadmap towards a modern simulation experience, watch Rescale’s presentation here.

Rescale Customer Panel: Innovator Perspectives on CAE Workflows

Rescale also hosted an engaging panel discussion focused on “The Future of CAE Workflow as told by Industry Innovators”, providing real-world perspectives on the trends many came to learn about. The panel featured seasoned modeling and simulation engineers including:

• George Oates, Thermal Lead Engineer at Rimac Technology: A leader in high-performance electric vehicles and EV technology, developing key components for large OEM’s hybrid and electric cars.
• Matthew Chung, Principal Engineer – CAE at Subsea7: A global leader in offshore projects and services for the evolving energy industry, utilizing HPC resources for complex subsea engineering.
• Novid Beheshti, Engineering Team Lead – CAE, CFD at Intelligent Energy: A hydrogen fuel cell manufacturer, serving sectors from automotive to aerospace with modular fuel cell products.

The panelists shared their experiences and insights into the evolving demands of CAE workflows. George Oates from Rimac Technology discussed the increased complexity of simulations driven by regulations, such as those related to battery thermal runaway, and the necessity of cloud computing to handle these demands. He anticipated a greater shift towards GPU-enabled workflows in the future. Matthew Chung of Subsea7 highlighted their successful use of cloud-based simulation to improve offshore operations, significantly reducing weather-related delays and saving millions of dollars. He emphasized how data-driven approaches and AI can enhance decision-making in time-sensitive operations. Novid Beheshti from Intelligent Energy spoke about their rapid migration to the cloud due to unpredictable compute peaks and the need for access to the latest hardware, including expensive GPUs. All three panelists envision AI automating repetitive and manual simulation tasks, freeing engineers to focus on more complex multi-physics problems and explore a wider range of conditions and design possibilities.

Key challenges and solutions discussed by the panel included data governance and storage, with the growing volume of data necessitating robust archiving systems and strategies to manage costs, such as enforcing data deletion policies. They also touched upon cloud adoption barriers, particularly in regions with strict data regulations like Germany, due to concerns about data security and export control. Rescale addresses these with features like government clouds and role-based access controls. A significant point of discussion was software licensing models; traditional licenses with core limits can hinder the scalability benefits of cloud computing. The panelists noted the value of “on-demand” or consumption-based licensing models, which align better with the elastic nature of cloud resources.

A crucial question from the audience addressed whether cloud computing is a good choice for enterprises, especially in Europe, given data residency and security concerns. The panelists’ response was an emphatic “yes.” Matthew Chung of Subsea7 stated that on-prem computing has its limits, but “on the cloud, it just unlocks all the potential with infinite processors.” Novid Beheshti from Intelligent Energy added that being on the cloud “helps massively” with unpredictable high peaks in compute demand. George Oates from Rimac Technology echoed this, emphasizing the elasticity of the cloud to have all the capacity needed and then turn it off, avoiding continuous hardware purchases for cyclical demands. This strong endorsement from industry leaders underscored the tangible benefits and growing confidence in cloud adoption for even the most sensitive engineering workloads.

Overview of Other Rescale Sessions

Rescale also contributed to several other deep-dive sessions at NAFEMs, showcasing new innovations and methodologies in cloud, AI, and HPC. These presentations delved into specific technical aspects and practical strategies for modernizing engineering workflows.

• Accelerating Scientific Workflows with Domain-Specific Hardware: GPUs, Arm Chips and Beyond by Dr. Sam Zakrzewski PhD and Romain Klein: This session explored how to optimize workloads with GPUs, ARM, and other accelerators using Rescale’s intelligent HPC platform, enabling unparalleled performance and efficiency for specialized applications.
• Challenges and Opportunities in Cloud-Based Simulation – An Engineer’s Perspective by John William and Romain Klein: This paper provided practical strategies for engineers migrating to cloud-based HPC, covering hardware selection, data handling, cost optimization, and license management.
• Safeguarding Engineering IP in the Cloud: Strategies for Secure Global Collaboration by Navin Bagga, Msc: This presentation focused on protecting sensitive intellectual property (IP) while enabling cross-border collaboration in the cloud, detailing strategies for managing end-user permissions and ensuring compliance.
• Orchestrating Hybrid HPC Environments: Strategies for Data Gravity and AI-Ready Datasets by Romain Klein and Carlos Mecha: This session delved into managing hybrid HPC setups and ensuring smooth data flow across on-prem and cloud environments, with a focus on tiered storage strategies and the Data Lake Exporter solution.
• Digital Thread Foundations for Accelerated Multi-Disciplinary CAE Workflows by Madhu Vellakal and Garrett VanLee: This paper discussed streamlining CAE with digital thread frameworks, AI-enhanced modeling, and real-world case studies, emphasizing the importance of centralized data structures and automated processes.
• Reshaping Simulation Data for an AI Future by John William and Romain Klein: This presentation proposed a comprehensive framework for simulation data management tailored to harness the power of AI for engineering applications, focusing on centralized data repositories, AI-powered analytics, and digital twins.
• Leveraging LLMs for Automated Post-Processing of Simulation Output Logs by James Imrie: This session explored the innovative use of Large Language Models (LLMs) in post-processing simulation software output logs to extract actionable insights efficiently, automate error classification, and build AI-ready datasets.

Download the ebook of Rescale session papers to explore detailed technical insights and strategies for optimizing your simulation workflows. 

Bring the Learnings to Your Organizations

The NAFEMS World Congress 2025 highlighted a clear industry consensus: cloud HPC, advanced data management (tackling the reality that data is the foundational problem to solve) and integrated AI (moving beyond chatbots to agents embedded in development loops) are fast becoming essential pillars of modern engineering. Rescale’s presence at the event—featuring Jacob Surber’s presentation and a dynamic customer panel—reflected how Rescale’s product roadmap aims to align with this future. The event as a whole reaffirmed the critical and evolving role of modeling and simulation in driving innovation across industries.

The insights shared by industry leaders from Rimac Technology, Subsea7, and Intelligent Energy provided proof of the promise of embracing these modern approaches, particularly the scalability and agility of the cloud and the transformative power of AI-enhanced simulation like the Rescale Assistant. Despite challenges related to data governance and licensing, the collective sentiment was a strong endorsement for the continued shift towards cloud-based, AI-augmented simulation. Notably, there was significant interest in leveraging data and AI layers independently, suggesting a growing recognition of these pillars as standalone value propositions.

To delve deeper into the innovations presented by Rescale at NAFEMS World Congress 2025, we encourage you to:

• Watch the full Rescale presentation with Rescale’s vision of modern simulation and the industry innovator panel to learn how Rescale is evolving the modern simulation experience.
• Sign up for our upcoming webinar for an overview and demos of Rescale’s new capabilities for data management and AI-assisted simulation.
• Contact us for any questions about how to accelerate adoption of the latest technologies for simulation, modeling, and AI for engineering.