Building Beyond Imagination: Engineering the Unthinkable with AI
Interview with Nico Haag of PhysicsX - CDFAM Amsterdam 2025
As simulation and design processes grow in complexity, the need for radically faster, more adaptive tools becomes increasingly urgent.
PhysicsX, a deeptech company founded by veterans of Formula One and the automotive sector, is developing AI-driven platforms that allow engineers to model and optimize physical systems in real time.
Bridging the gap between high-fidelity numerical simulation and practical design application, PhysicsX works across aerospace, energy, automotive, and advanced manufacturing—embedding large physics models (LPMs) directly into engineering workflows.
Their hybrid approach combines deep expertise in simulation and machine learning with hands-on collaboration to integrate tools within real-world product development cycles.
In this interview, co-founder and Director of Engineering Nico Haag outlines the origins of PhysicsX, the foundational challenges that led them to adopt AI in simulation, and how the company is scaling its platform from early concept through to manufacturing integration.
We also discuss the company’s public demonstrator Ai.rplane, how LPMs are helping reduce design timelines from years to months, and what they hope to bring to—and take away from—the CDFAM Computational Design Symposium in Amsterdam.
To start, can you give an overview of PhysicsX? What does the company do, and how do you balance software development with consulting services?
PhysicsX is a deeptech company with a focus on accelerating hardware innovation at the speed of software. Our roots are in numerical physics and Formula One, but our impact spans advanced industries like Aerospace & Defense, Energy, Materials, Semiconductors, and Automotive.
At our core, we’re building an AI-driven simulation software stack that enables high-fidelity, multi-physics simulations, bringing a level of speed and automation to engineering that was previously unimaginable.
Engineering simulations are essential for developing complex machines and processes, from designing fusion reactors to manufacturing jet engines. However, current simulation technologies are challenging to set up and computationally expensive to run. PhysicsX is pioneering AI-driven methods to drastically speed up physics simulations and accelerate innovation, enabling engineers to design, test, and refine configurations in real time. This is made possible through novel techniques for generating 3D geometries, predicting their dynamic behavior, and optimizing performance.
What sets us apart is our hybrid approach and the depth of our in-house expertise. With over 120 simulation, machine learning, and software engineers, we don’t just develop off-the-shelf tools and leave our clients to figure out how to best use them — we work directly with the organizations to deploy these tools in a way that maximizes impact and addresses existing challenges. That’s where the balance between software development and consulting comes in.
Our AI-driven platform is at the heart of what we do, but real impact comes from embedding that technology into engineering teams and workflows. We collaborate closely with customers to pinpoint their most complex challenges, accelerate their simulation capabilities, and ultimately push the boundaries of what’s possible in design, manufacturing, and operations. It’s not just about delivering software; it’s about ensuring that software delivers real-world, measurable advancements in engineering.
How did PhysicsX get started? What was the first engineering challenge that led to exploring AI-driven solutions, and how did that shape the company’s direction?
PhysicsX was founded with a clear mission: to bring the power of AI to high-fidelity physics simulation and accelerate engineering innovation. The idea came from firsthand experience in Formula One, the fastest development cycle in engineering on Earth, where the need for ultra-fast, high-accuracy simulations is critical. Traditional simulation methods, like computational fluid dynamics (CFD) and finite element analysis (FEA), while powerful, can set teams back due to computational cost and time constraints. We saw an opportunity to break that bottleneck using AI.
The first challenge we tackled was improving aerodynamic simulations — specifically, accelerating computational fluid dynamics (CFD) without sacrificing accuracy. By applying AI to predict high-fidelity physics outcomes in real-time, we proved that engineering workflows could be radically optimized. That early success set the foundation for our broader approach: developing an AI-driven simulation stack that works across multiple physics domains and industries, from Aerospace to Semiconductors.
One of your software projects, Ai.rplane, serves as a demonstration of your approach. Can you explain what it does and how it reflects the broader capabilities of PhysicsX?
Ai.rplane is a perfect example of how we apply AI to high-fidelity physics simulation. It’s an interactive technology demonstrator that allows engineers to explore real-time aerodynamic responses without waiting for traditional CFD computations. Rather than running full physics simulations for every parameter change, Ai.rplane uses our Large Geometry Model, LGM-Aero, to provide instant, high-accuracy results — enabling broader design exploration and unparalleled optimization capabilities.
This reflects the broader vision of PhysicsX: embedding AI-driven simulation into engineering workflows to make high-fidelity modeling orders of magnitude faster and more accessible. Ai.rplane is a demonstration of what’s possible with PhysicsX — not just in aerospace, but across industries where simulation speed and cost are limiting factors. It’s about removing barriers to rapid iteration, enabling engineers to make data-driven decisions in real time.
For companies considering AI-driven design, how do they know when it’s the right time to engage with PhysicsX? What types of challenges or goals make them a good fit for your platform?
The best time to engage with PhysicsX for companies that operate in the hardware design and manufacturing space is now.
If teams are running thousands of expensive, time-consuming simulations or relying on oversimplified models, leaving significant performance gains on the table. In Formula One, teams win by out-innovating their competition. It should be the same in industry — applied AI is becoming the backbone of modern business, driving change and turning what was once experimental into the essential.
Companies that can gain a competitive advantage with our platform typically have:
– Complex, high-fidelity simulation needs (CFD, structural, thermal, multi-physics, etc.).
– Large design spaces where faster iteration leads to better outcomes.
– A desire to move beyond traditional parametric optimizations to AI-driven generative design.
Ultimately, PhysicsX delivers real value to engineering teams seeking to push their designs further and faster without sacrificing accuracy while also optimizing manufacturing and operational processes.
A common barrier to AI adoption is data availability. What kind of data does a company need to bring to the table, and how long does it typically take for them to become independent users of your platform?
Yes its a common concern, but our platform is designed to tackle that head-on.
We can ingest a wide variety of data sources, including simulation data, physical measurements, and process data, ensuring companies can leverage whatever information they have available, if any.
Our (pre-trained) models are very data-efficient, meaning they extract maximum value from limited datasets, reducing the need for massive historical databases. This allows companies to achieve reliable, high-performance AI models faster than traditional approaches.
Additionally, we have a large team of brilliant engineers that (1) deeply understand the engineering challenges and (2) can also create high-quality simulation data, enabling accurate, reliable models even in data-scarce environments. Additionally, our platform includes streamlined workflows for complex simulation processes, ensuring consistency and minimizing manual effort. This combination empowers companies to quickly ramp up and become independent, productive users of our platform — without waiting for years of data collection.
In terms of timeline, we typically see meaningful results within weeks.
Full deployment and independent usage depend on the complexity of the problems an organization is looking to solve, but we design our tools to integrate seamlessly with existing workflows, allowing teams across the enterprise to adopt them progressively without major barriers.
At CDFAM, you’ll be presenting on Large Physics Models (LPMs) and their role in AI-driven engineering. What are the key takeaways you hope attendees will gain from your presentation?
The key message is that Large Geometry Models (LGMs) and Large Physics Models (LPMs) are fundamentally changing how engineering is done. Just as Large Language Models (LLMs) transformed natural language processing, LGMs & PMs are revolutionizing physics-based design and optimization.
Attendees should walk away with a clear understanding of:
– Why LGMs & LPMs matter: how they enable designs beyond human imagination at orders-of-magnitude physics acceleration to drastically reduce time-to-market and reach ultimate performance while reducing cost to a minimum.
– How they integrate into engineering workflows, bringing AI-driven predictions to real-world applications.
– What’s next: how these models will drive the next generation of automated design and manufacturing.
Ultimately, we want engineers to see LPMs not as a future concept, but as a practical, deployable tool that can dramatically enhance their capabilities today.
Beyond sharing your work, what are you hoping to gain from participating in CDFAM? Are there specific collaborations, insights, or discussions you’re looking forward to?
CDFAM brings together some of the best minds in computational design and advanced manufacturing, so for us, it’s about meaningful collaboration.
We’re particularly interested in discussions around the intersection of AI and physics-based simulation — how others are tackling similar challenges, where the biggest bottlenecks remain, and how we can push this technology even further.
We’re also looking to connect with engineering teams who are ready to adopt AI-driven simulation but need help figuring out the path forward. It’s always exciting to hear firsthand about real-world challenges and explore how our technology can make a tangible impact.
For me, CDFAM is about advancing the field together. We’re bringing cutting-edge solutions to the table, but we’re just as eager to learn from others who are pushing the boundaries of what’s possible in design and manufacturing.
The CDFAM Amsterdam 2025 program brings together leading experts in computational design from industry, academia and software development across all scales of application, from micro to mega.
This year’s keynotes will be delivered by Federico Casalegno, Executive Vice President of Design at Samsung Electronics, and Mathew Vola, Arup Fellow of Computational Design alongside over 30 experts like Nico from around the world.
Space is limited, so register now to secure your seat at the world’s leading computational design and engineering event.