JLR is redefining manufacturing by integrating AI and digital twins to enhance efficiency, optimise production, and drive innovation. Mark Venables explores how Paulina Chmielarz is leading this transformation, leveraging real-time data, predictive analytics, and generative AI to reshape industrial operations.
Paulina Chmielarz has spent years at the heart of industrial operations, shaping the digital transformation of JLR’s supply chain, manufacturing, and procurement processes. She is now the Industrial Operations Digital and Innovation Director at JLR, and speaking at NVIDIA GTC, she outlined how JLR has embraced digital twins and AI to redefine its approach to factory launches, data integration, and operational efficiency.
Rethinking factory launches with AI
Launching a new automotive factory is a complex and high-stakes endeavour. Thousands of interconnected elements, new products, processes, and new employees must align seamlessly to bring a facility online. Chmielarz has led such launches, understanding firsthand the challenges of integrating disparate systems, maintaining data consistency, and ensuring cross-functional connectivity.
Years later, at NVIDIA’s headquarters, she encountered technology that could fundamentally transform this process. Digital twins offered the ability to relive past events, rewind time, and analyse operations with unprecedented clarity. It was a compelling opportunity: “Managing and consuming data remains one of the biggest challenges in any industry,” she says. “Compatibility is a constant struggle, as large IT systems speak different languages and operate on different data models.” Aligning planning systems and execution layers, each running on different time cycles, created further barriers. The solution lay in a new way of structuring data and decision-making.
Rather than beginning with theoretical models, JLR’s approach prioritised integrating real-world IoT data. “Bringing live operational data into the digital twin provided a stronger foundation for every subsequent stage,” Chmielarz adds. Instead of a linear approach, planning, validation, and integration, JLR flipped the model, using real-time insights to drive operational improvements at every step. This allowed for a more iterative approach, refining systems dynamically based on real-world conditions rather than static assumptions.
Building a new data foundation
Digital transformation is not just about technology; it is about people. The first step was building a team of coders, VR developers, specialists, and game designers to explore the Omniverse. “We started as a small team and expanded from four to ten, then to fifteen and twenty-five, supported by subject matter experts,” Chmielarz continues. The team could move beyond isolated projects to systemic change with a foundation.
Generative AI was introduced to enhance this system. JLR implemented AI-powered chatbots and headless connectors to increase data accessibility, allowing teams to retrieve historical data efficiently. “Tracing performance issues back to their root causes weeks later was a significant breakthrough,” Chmielarz says. “The ability to analyse past events with high fidelity enabled engineers to pinpoint inefficiencies and refine operations before problems escalated.”
The team also prioritised an interactive user experience to ensure accessibility. Cloud-based remote rendering made the system mobile, while predictive quality tools and production optimisation enhanced decision-making. Integrating over 100 IoT data streams into a unified model set the stage for a more structured validation process.
With a growing volume of data, JLR focused on enhancing cybersecurity within its digital twin ecosystem. Protecting sensitive operational data from cyber threats requires a combination of encryption, secure access controls, and AI-driven anomaly detection. “Ensuring the integrity of our data was critical. AI allowed us to detect potential security breaches before they could impact production proactively,” Chmielarz says. “By incorporating AI-driven cybersecurity measures, we strengthened its digital twin infrastructure, ensuring reliability and resilience.”
Another key challenge was ensuring interoperability across different digital platforms. “The automotive industry relies on a mix of legacy systems and cutting-edge technology, often resulting in data silos,” Chmielarz explains. “Bridging these gaps is essential. We need platforms that can communicate seamlessly with one another. By adopting open standards and scalable data architectures, we reduced inefficiencies caused by fragmented systems.”
From validation to real-time simulation
Validating data before implementation is essential in manufacturing, where even minor adjustments can disrupt production. Digital twins allowed JLR to simulate changes before deploying them on the factory floor. “When modifying an ecosystem with thousands of interdependent elements, rigorous validation is essential,” Chmielarz says. “Real-time validation provided a controlled environment to test, refine, and optimise before committing to large-scale changes.”
Beyond validation, JLR leveraged synthetic data to train AI models. Using Omniverse Replicator, the team generated 12,000 annotated images daily to improve production line algorithms. By mirroring real-world scenarios, engineers could conduct detailed root-cause analyses and track operational performance more accurately. With multiple data streams integrated through Universal Scene Description (USD), the digital twin evolved into an interactive, high-fidelity model. This framework enabled JLR to reconstruct past events precisely, offering a more detailed view of production processes.
Simulation was not limited to past events. JLR leveraged predictive analytics to model future scenarios, optimising everything from production scheduling to supply chain logistics. “By forecasting demand fluctuations and adjusting supply chain strategies accordingly, we improved overall efficiency and reduced waste,” Chmielarz adds. “AI-driven predictions enabled JLR to identify potential bottlenecks before they materialised, preventing costly disruptions.”
Optimising manufacturing with AI
The benefits of digital twins were particularly evident in the body shop, where high-precision robots weld sheet metal panels to form the car’s structure. “This process demands extreme accuracy, making it an ideal area to apply our technology,” Chmielarz explains. “We optimise robotic performance by integrating real-world data sources, enhancing speed and precision.
“To maximise simulation efficiency, unnecessary visual elements were removed from Omniverse models. We focused solely on performance, allowing us to run high-load simulations without overloading the system.” This approach improved dynamic visualisation, enabling engineers to interact with data in new ways and refine processes in real-time.
Cloud XR further accelerated system performance while integrating NVIDIA NIMs, which improved language processing for AI-powered workflows. These advancements turned the digital twin into an enterprise-wide tool, not just a technical proof of concept but a fully operational system capable of driving tangible business outcomes. JLR also explored AI-assisted human-robot collaboration, integrating AI-powered vision systems to enhance quality control. “Automating quality inspections improved consistency while allowing human workers to focus on more complex problem-solving tasks,” Chmielarz adds. “AI-enhanced quality control reduced defects and rework, contributing to a more efficient production process.
“Another area of focus was workforce augmentation through AI-driven guidance systems. By equipping our teams with augmented reality tools and AI-based digital assistants, we enhanced their ability to execute complex assembly tasks more accurately. These systems provided step-by-step guidance and real-time alerts, reducing errors and improving training efficiency.”
The future of digital operations
Despite the advancements in AI-driven manufacturing, some things remain unchanged. “People still take screenshots and send them via email,” Chmielarz notes, highlighting the persistence of familiar workflows even in cutting-edge environments. This anecdote serves as a reminder that technology adoption is as much about behaviour as it is about capability.
Reflecting on JLR’s journey, three key lessons stand out. First, innovation requires freedom to explore; business outcomes should not dictate early experimentation. Learning must come first. Second, expertise is essential; technical knowledge cannot be underestimated. Finally, technology should not just improve efficiency but also engage and inspire. “As one of our factory managers put it, this technology is not just powerful; it is fascinating,” Chmielarz says.
JLR’s approach to digital transformation is a testament to the power of AI and digital twins in modern manufacturing. By rethinking how data is integrated, validated, and simulated, the company has optimised operations and set a precedent for the future of industrial digitalisation. The journey is ongoing, but the impact is already tangible.
Looking ahead, JLR is investigating AI-driven sustainability solutions. “Reducing energy consumption and optimising resource use is a key focus area,” Chmielarz concludes. “AI allows us to analyse and improve energy efficiency across our production network.” By leveraging AI for predictive energy management, JLR aims to reduce its environmental impact while maintaining production efficiency. AI’s role in sustainability extends beyond energy efficiency, optimised logistics and smarter resource allocation are becoming integral to future-ready manufacturing operations.
