AI/ML for Science (1): AI in Engineering Product Lifecycle and Enterprise Software
How AI is Transforming Engineering: From Design and Simulation to Testing and Maintenance
AI is Eating Enterprise Software - But What About Engineering?
As I explore AI/ML in the enterprise software space, I’m constantly amazed by how many innovators and entrepreneurs are finding new ways to integrate large language models (LLMs) and AI-driven systems into products. Startups were racing to integrate AI copilots into every workflow, automating emails, coding, and marketing. But Engineering AI was playing by a completely different set of rules.
In consumer AI, fast iteration and viral adoption drive success. If a chatbot messes up, it’s just annoying. But in engineering, bad AI can have serious consequences. Image if AI miscalculates an autonomous driving decision or gives the wrong answer in a material strength test, these decisions can lead to serious, expensive failures or even safety risks.
AI adoption in engineering can take slower but deeper steps. Instead of just automating simple tasks, AI is being built into the core of how products are designed, tested, and maintained. And while big players like Tesla, Boeing, Siemens, GE are leading the charge, I’ve been seeing a wave of startups quietly pushing AI into some really niche, but really powerful applications.
In last blog, I explored How AI is Revolutionizing Academic Research, focusing on AI-powered R&D tools. In this article, I’ll dive into AI-driven solutions in science and engineering, exploring how AI is transforming design, simulation, testing, and maintenance—and how researchers, industry leaders, and startups are shaping this space.
In the research space, I found this DOE report particularly insightful to understand advanced research directions in AI for science and energy. On the industry side, I have seen engineering leaders and engineering software companies actively funding the researches efforts to reduce cost and improve workflows through AI-driven solutions. At the same time, researchers in academic and corporate R&D are commercializing their breakthroughs, bringing cutting-edge innovations to the capital market and driving real-world adoption through collaborating with incumbents
AI in Engineering Design: Two different Challenges and Solutions
1. AI-based Design
One of the shift we seen in engineering design is that AI-driven generative design. Instead of an engineer manually tweaking designs, AI suggests thousands of optimized structures based on real-world constraints, create designs and integrate physics into CAD workflows.
AI-powered generative design is evolving fast, but there’s still a gap between what’s possible in research and what’s ready for industry use. Some cutting-edge research areas include:
Deep Learning for Mechanical Design: Researchers at MIT and Berkeley are exploring physics-aware deep learning models that generate designs not just based on geometry, but also on material properties and real-world constraints.
AI + Topology Optimization: Stanford researchers are combining AI with topology optimization, helping AI design parts that are both lighter and stronger—critical for aerospace and automotive.
Who’s Leading the Charge?
Autodesk Fusion 360 is a design, engineering, electronics, and manufacturing all-in-one software. It is now using artificial intelligence algorithms to generate and optimize designs. It uses algorithms and machine learning techniques to explore a range of design possibilities to find optimal solutions based on given constraints and objectives. Their users now can input design goals and constraints, and the AI system generates multiple design options that meet those requirements.
Siemens NX is a leader in advanced high-end CAD/CAM/CAE. It enable users to leverage their own IP for training machine learning models within a secure environment. It also adopted AI to enhance user experience by helping eliminate redundant tasks such as reducing clicks and time spent on routine tasks. This allows users dedicate more time to innovating and help reduce product development lifecycles.
In the startup space, last stage growth company nTop’s platform uses AI-driven generative design and topology optimization. Their platform allows engineers to create highly optimized, lightweight, and complex structures that traditional CAD tools struggle to handle.
nTop Workflow: Build computational models, Run computational models, Integrate with downstream tools
2. AI and Surrogate Modeling for Faster and Cheaper Computation in Design
Now, let’s talk about the biggest bottleneck in engineering design - physics simulations. When building complex engineering products like car or airplanes, ever touches the real world, engineers need to simulate real-world conditions ot know how it will perform. For example, predicting how air flows over it, how strong its materials are, and how it will hold up under different conditions.
What’s the problem?
Traditionally, answering these questions requires massive, high-fidelity simulations using Computational Fluid Dynamics (CFD) for aerodynamics or Finite Element Analysis (FEA) for structural integrity. These simulations are slow and expensive, sometimes taking days or even weeks to run on powerful supercomputers. And when designs change (which they often do), the process starts all over again.
Instead of generating the design, AI steps in by transforming the way we model complex physical systems. AI-driven surrogate models act as shortcuts, learning patterns from past simulations to generate accurate predictions in a fraction of the time. These AI models combine low-cost, lower-resolution simulations with a smaller amount of high-accuracy data, speeding up the design process without sacrificing precision.
Researchers are actively developing this technology for real-world applications:
Automotive: MIT and Toyota Research Institute trained an AI model to use simplified 2D representations instead of full 3D simulations. Their AI can now predict a car’s drag coefficient—a key measure of aerodynamic efficiency—quickly and accurately, cutting down expensive simulation time.
Aerospace: Researchers in Stanford developed an AI-driven surrogate model that combines low-fidelity simulations with a small set of high-fidelity data, allowing AI to predict aerodynamic forces, drag, and lift coefficients much faster. This means aircraft designers can test and optimize wing shapes in hours instead of weeks, leading to faster innovation in fuel-efficient aircraft, drones, and next-gen supersonic jets.
A Swiss startup Neural Concept developed a platform that uses 3D deep learning for aerodynamic modeling to optimize component geometry and material choices by learning from simulated and tested data while automating routine CAD tasks based on successful past projects.
London-based PhysicsX just came out of stealth. They came up with an AI platform to create and run simulations for engineers working on project areas like automotive, aerospace and materials science manufacturing — industries where there are regularly bottlenecks in development due to how models are tested before production.
Testing and Quality Assurance: Smarter, Faster, and More Reliable
In the previous section, I talked about optimization in the design phrase. Finding and fixing design flaws in testing is also a big part before mass production. Testing is usually time-consuming, expensive, and sometimes impractical. Think about stress testing an aircraft wing—it involves physically applying force until failure, which takes time, money, and multiple prototypes. AI is changing the game by bringing predictive insights and automated testing into the process.
How AI is transforming testing and QA?
Instead of relying on human inspectors to spot flaws, AI-powered computer vision can scan thousands of parts per minute, detecting microscopic defects in car bodies, airplane fuselages, and semiconductor chips. AI can predict material stress and failure points before real-world tests even happen.
Engineers are also working on developing a digital twin for Virtual Prototyping. A Digital Twin is a real-time AI-powered virtual replica of a product—whether it’s an airplane engine, a factory assembly line, or a nuclear reactor. AI continuously learns from real-world sensor data to predict failures before they happen.
Tesla and BMW use AI-powered QA systems that analyze images from factory production lines, flagging defects in real-time before faulty parts make it into vehicles.
Akselos is a Swiss based startup grew out of the MIT research group. It designs digital twins of the world’s most complex systems using AI-powered structure simulation technology to help guard critical infrastructure by creating a virtual replica of an asset in its current environment, monitor system condition in real-time and predict failures before they happen.
UK-based Series A startup Monolith AI provides a self-learning model based on existing test data and replace trial-and-error with predictive modeling. For example, it helped BMW group to use the wealth of their existing crash data to optimize crash performance earlier in the design process and reduce dependence on time-intensive, costly testing.
Applied Intuition offers AI-powered simulation and testing tools for autonomous vehicles, ensuring self-driving systems are robust before real-world deployment. Its development platform allows engineering teams to safely develop and test ADAS and AD systems at scale. It uses machine learning (ML) techniques to support increasingly realistic and complex use cases.
By making testing faster, smarter, and more automated, AI is reducing waste, improving safety, and helping companies build better products in less time.
AI in Predictive Maintenance: Fixing Problems Before They Happen
Lastly, I want to talk about maintenance part in engineering companies. Imagine a factory running 24/7, filled with robotic arms, conveyor belts, and heavy machinery. If one part fails unexpectedly, production stops, causing delays, lost revenue, and expensive repairs. Companies need to schedule maintenance on a fixed timeline, replacing parts whether they’re worn out or not. Machine Learning and AI transform equipment management in maintenance and operations by predicting failures before they happen. More specifically, AI can analyze sensor data to predict failures and optimize service intervals. This predictive maintenance approach helps engineers identify efficiency improvements and energy savings while uncovering patterns in failure data to prevent issues from recurring.
How AI is improving maintenance right now?
The rise of Industrial IoT (IIoT) and AI-powered predictive analytics is enabling factories, power plants, and supply chains to move from reactive to proactive maintenance, saving millions in downtime costs.
GE & Siemens use AI-powered sensor monitoring to analyze vibration patterns, heat signatures, and electrical signals from industrial machines, flagging early signs of failure before breakdowns occur.
Tesla’s Gigafactories uses AI-driven predictive maintenance systems optimize battery production, preventing breakdowns in high-speed manufacturing lines.
Series C company Altana developed a enterprise platform helps company manage global supply chain all the way to upstream raw materials as well as customer network. The system also uses AI to provide insights and recommend actions. They are collaborating with Maersk to use AI to monitor large cargo ships, predicting mechanical failures before they reach open waters, reducing fuel consumption and engine failures.
Instead of fixing things when they break, AI enables proactive maintenance, keeping factories, airplanes, and even power plants running smoothly—saving time, money, and resources.
Final Thoughts: AI is Reshaping Engineering
Across engineering product lifecycles, AI is revolutionizing workflows by providing more efficient design and simulation to increase the test success rate, improving testing and quality assurance for safer and faster product to market, predictive operation and maintenance in further reducing cost, and there are more use cases!
Will AI replace engineering jobs? No. Just like GenAI serves as a personal assistant in daily tasks, AI can handle repetitive, time-consuming tasks, but it doesn’t have the creativity, domain expertise, or intuition that human engineers bring to product innovation. While AI can suggest better designs and automate simulations, it’s always a human who makes the final decision. Engineers will still drive innovation, but now with AI-powered tools to make their work faster, smarter, and more efficient.