Autonomous vehicles rely on more than just sensors to navigate safely. They need high-definition (HD) maps that provide real-time environmental data. AI-driven mapping revolutionizes how these maps are built, maintained, and updated, ensuring that self-driving technology can operate effectively. By leveraging AI, computer vision, and MLOps, companies push the boundaries of what’s possible in autonomous navigation.
To dive into this topic, we spoke with Julia Flament-Wallin, an expert in applied AI and software engineering, and one of the speakers at the Data Innovation Summit 2025! In this interview, Julia shares how AI-powered mapping enhances autonomous driving, the role of computer vision in detecting map features, and the importance of real-time change detection. She also discusses the critical role of MLOps and platform engineering in scaling HD map automation.
Read on to learn how AI is reshaping the future of HD mapping and what it means for the next generation of autonomous vehicles!
Hyperight: Julia, can you tell us about your journey into applied AI and software engineering? What inspired your interest in mapping and autonomous driving?
Julia Flament-Wallin: I’ve had one leg in data and one in engineering since I started my professional career. I have always been drawn to the back-end system design. Maps, on paper or my computer, are something I’ve used and enjoyed for as long as I can remember. I always prefer a map over a list view. I often miss clearly shown elevation, as I have a kind of a phobia of holes. I like the view. Anyway, when the chance came to work with TomTom building maps by applying engineering and AI in combination, I didn’t need much convincing!
Hyperight: Your presentation at the Data Innovation Summit 2025 is focused on AI-powered mapping for autonomous driving. What can the delegates expect from your presentation?
Julia Flament-Wallin: I hope to be able to lift the lid on how we automate map building, which is how we’re able to do it at scale, and why that is so crucial – especially for autonomous driving. Which is still an unsolved problem except for a few restricted areas.
Hyperight: Julia, can you elaborate on the role of computer vision and AI in detecting and classifying features for HD maps? How do these technologies complement traditional mapping methods?
Julia Flament-Wallin: A collection of algorithms, that fall under Computer Vision and AI, is how we extract the features we need to create HD Maps from the various image sources we rely on. Be that aerial images or sensor data from cars. These sources are what enable modern map automation.
Hyperight: You will mention MLOps and platform engineering in your talk. How do these practices help streamline the process of creating HD maps?
Julia Flament-Wallin: MLOps is crucial as soon as you want to have 1 or more ML models in production. For example, we need to know that we’re not introducing any regression to the map by deploying new or improved versions of models. Platform engineering comes into play for us because we publish our features to the TomTom platform, Orbis. As Value Streams or (map) features teams, we’re not building the platform per se, but we need to implement integration with it in every way that we can, which is a very collaborative process with the platform team.
Hyperight: Why are HD maps still crucial in autonomous driving, and how do they enhance sensor data?
Julia Flament-Wallin: This is an extremely competitive field, and time will tell which approach “wins.” Some think it will be SD maps + sensors. Making HD maps is not trivial, but obviously, I’m biased towards TomTom’s strategy.
Hyperight: Change detection is a critical aspect of HD mapping. Can you explain how AI models detect and handle changes in dynamic environments?
Julia Flament-Wallin: The world changes as we’re building a model to represent it – the map. A road changes direction, a sign no longer exists, a new one pops up, etc. We get new data all the time, and use different algorithms to be able to detect only changes in data.
Hyperight: Looking ahead, what advancements or trends do you foresee in the intersection of AI, mapping, and autonomous driving in the upcoming years?
Julia Flament-Wallin: I’m not a huge fan of making predictions, but the demand for location technology is huge. I think that as soon as map makers, autonomous driving systems and location technology will become even more mature in the coming years, so will expectations from customers.
If you want to learn more about the future of AI-driven HD mapping and its role in autonomous driving, don’t miss Julia’s session at the Data Innovation Summit 2025! She will break down the complexities of AI-powered map automation, explain how computer vision enhances HD maps, and discuss the impact of MLOps in scaling these technologies.
Join Julia as she explores the challenges and breakthroughs in autonomous navigation. Whether you’re interested in AI, mapping, or the future of self-driving vehicles, her session will provide insights into these innovations shaping the road ahead!
Add comment