Master Thesis Program 2027, Cluster B: Data, 3D Reconstruction & Generative AI

Help build the foundations behind modern AI What if your master’s thesis could help AI systems understand and learn from the real world more accurately? At Zenseact, advanced AI depends on more than neural networks. It requires high-quality data, reliable sensors, realistic 3D environments, scalable annotation methods, and efficient tools for experimentation. Cluster B brings together AI, computer vision, 3D reconstruction, generative models, sensor technology, and data engineering. You will work on a defined research challenge using relevant data and modern methods, contributing knowledge that may support safer and more capable driving technology. You do not need to know the answer before you begin. What matters is your ability to investigate a problem thoughtfully, learn from evidence, and develop your understanding throughout the thesis. Choose Your Project B1: 3D Gaussian Splatting Reconstruction for Enhanced Parking Explore how 3D Gaussian Splatting can reconstruct detailed driving environments and support enhanced parking applications. B2: Data Curation with Self-Supervised Vision Foundation Models Investigate how foundation models can organize and curate autonomous-driving data at scale, helping identify relevant scenes without relying entirely on manual labels. B3: Multi-Camera LiDAR Calibration via 3D Gaussian Splatting Develop and evaluate methods for jointly calibrating multiple cameras and LiDAR using a shared 3D representation. B4: Weather Editing for Occupancy-Guided World Models Explore how weather conditions can be modified in driving scenes while preserving the geometry needed for reliable world models and evaluation. B5: Robust 3D Head Reconstruction for Digital Human Generation Investigate methods for reconstructing human heads in 3D and generating realistic digital human representations. B6: Diffusion-Based Image Synthesis from 3D Point Clouds ( Location for this project: Lund, Sweden) Explore how diffusion models can generate images from reconstructed 3D point clouds and how geometry can guide realistic image synthesis. Read more about the projects here: https://zenseact.com/mtp_cluster-b/ Your Mission Every Cluster B thesis combines research with practical experimentation. You will work with real-world data, develop and evaluate methods, analyze results, and contribute insights that may help shape future AI and autonomous-driving systems. You will have significant ownership of your work while receiving regular guidance from experienced engineers and researchers. Together, you will define a realistic scope, discuss technical choices, and adapt the direction as new evidence emerges. Depending on your chosen project, your work may include: Reviewing relevant scientific literature Preparing and analyzing image, LiDAR, point-cloud, or annotation data Designing and implementing experiments Training or adapting models and 3D representations Defining suitable baselines, metrics, and evaluation methods Investigating performance, limitations, and failure cases Documenting your work and presenting your conclusions Who You Are You are curious about how AI systems learn from and represent the physical world. You enjoy exploring challenging technical questions, testing ideas, and learning from results, including when they do not match your initial expectations. You are currently pursuing a master’s degree in Computer Science, Machine Learning, AI, Robotics, Electrical Engineering, Engineering Physics, Applied Mathematics, or a related field. We value different perspectives, experiences, and ways of approaching problems. You do not need to match every item below. We are interested in your technical foundation, learning ability, motivation, and potential to grow during the thesis. What You Bring Programming experience, preferably in Python Academic or practical experience in machine learning, deep learning, computer vision, data analysis, geometry, or a related area Ability to read and evaluate scientific literature Ability to communicate technical reasoning and results in English Eligibility to complete the thesis as part of a university-level course Your thesis must be approved by a university examiner and supervisor and completed as part of a university-level course. Helpful, but Not Required Depending on the project, experience in one or more of the following areas may be useful: Deep learning frameworks such as PyTorch Foundation models, self-supervised learning, or generative AI Computer vision, LiDAR, point clouds, or 3D perception Sensor calibration, neural rendering, or 3D Gaussian Splatting Data pipelines, GPU computing, or large-scale experimentation Research projects, publications, or open-source contributions Previous experience with the exact method named in your preferred project is not required. We encourage you to apply if you have relevant foundations and are motivated to learn. Why Cluster B? Cluster B is for students interested in the technical foundations that make modern AI development possible. You will explore technologies shaping the future of intelligent systems, from foundation models and generative AI to 3D reconstruction, sensor calibration, and large-scale data curation. At the same time, you will gain experience in turning an open research question into a structured investigation with meaningful results. Your thesis will connect academic research with real technical challenges inspired by autonomous driving. You will have the opportunity to deepen your expertise, learn from experienced colleagues, and contribute your own perspective in an international engineering environment. If you are excited by AI, computer vision, 3D understanding, generative models, or data-intensive machine learning, we would like to hear from you. Practical Information Thesis period: Spring 2027 Location: Gothenburg, Sweden (project B6 is in Lund) Format: Individual applicants and thesis pairs are welcome Working model: Primarily on-site collaboration with flexibility when app