Computer Vision & Perception for AD: Master Thesis Projects 2027, Cluster A
Confirmed on the employer's own hiring board on Oct 19, 2026. First seen by Alion on Sep 8, 2026.
Understanding the world is one of the fundamental challenges in artificial intelligence.
At Zenseact, we build AI systems that must understand, interpret, and predict complex real-world traffic environments. Our perception and world-modeling technologies are trained on large-scale driving data and contribute to future autonomous-driving capabilities.
Cluster A focuses on some of the most exciting areas in modern machine learning, including self-supervised learning, multimodal perception, world models, synthetic data, occupancy prediction, and 3D scene understanding.
As a thesis student, you will work on open-ended research challenges where the answers are not yet known, contributing ideas that may influence future generations of perception systems.
Choose your project:
A1: Self-Supervised Learning for Camera-Radar Perception in Autonomous Driving
Explore how self-supervised learning can improve multimodal perception by combining camera and radar data without relying heavily on manual labeling.
A2: Self-Supervised Learning via a Joint Embedding Predictive Architecture
Investigate predictive representation learning methods that allow models to learn directly from autonomous-driving imagery and large-scale unlabeled data.
A3: Enhance Dynamic Object State Prediction Using Synthetic Data
Study how synthetic data can improve prediction capabilities for dynamic agents and support more robust autonomous driving systems.
A4: Towards Better Occupancy World Models for Autonomous Driving
Develop new methods for understanding and predicting occupancy-based representations of the driving environment.
A5: 3D-Grounded World Models for Autonomous Driving
Explore how world models can incorporate 3D understanding and physical structure to better represent complex traffic environments.
A6: Semantic Superquadric Splatting for Structured 3D Scenes
Research novel approaches for semantic scene understanding using modern 3D representations and structured scene modeling.
Read more about the projects here: https://zenseact.com/mtp_cluster-a/
Your mission and day-to-day tasks
You will combine machine-learning research with hands-on experimentation using real sensor data and modern compute resources.
Together with experienced researchers and engineers, you will:
Review state-of-the-art research
Work with large-scale sensor datasets
Design and run deep-learning experiments
Train and evaluate neural networks
Build perception and world-model architectures
Analyze model behavior and failure cases
Present findings and recommendations
You will have significant ownership of your thesis while receiving guidance from experts in the field.
Qualifications & Experience
We're looking for students with strong technical foundations, curiosity, and a desire to learn.
You are currently pursuing a Master's degree in Computer Science, AI, Machine Learning, Robotics, Engineering Physics, Electrical Engineering, Applied Mathematics, or a related field.
Essential
Strong Python programming skills
Understanding of machine learning fundamentals
Experience working with data and experimentation
Ability to read scientific literature
Strong English communication skills
Helpful, but not required
Experience in one or more of the following areas:
Deep learning frameworks such as PyTorch or TensorFlow
Foundation models, self-supervised learning, or representation learning
Computer vision, sensor fusion, or world models
Occupancy prediction or 3D scene understanding
Research projects, publications, or open-source contributions
Previous experience with the exact methods used in a project is not required. Strong fundamentals, curiosity, and a willingness to learn matter most.
Why Cluster A?
Cluster A combines cutting-edge machine-learning research with real-world deployment challenges.
You will work on problems at the intersection of computer vision, representation learning, robotics, and autonomous driving, using large-scale driving datasets and modern AI infrastructure.
If you are excited by understanding how intelligent systems learn about the world, and want to contribute to technology that could improve road safety at scale, Cluster A is built for you.
Check out Cluster B and Cluster C aswell, before you apply.
Practical Information
Thesis period: Spring 2027
Location: Gothenburg, Sweden
Format: Individual applicants and thesis pairs are welcome
Working model: Primarily on-site collaboration with flexibility when appropriate
Eligibility: You must be enrolled in a Master's program and conduct the thesis as part of your university studies
Legal requirement: You must have the legal right to study in Sweden during the full thesis period
Application: Please rank your top three project choices when applying
This role may have access to sensitive information, trade secrets, and confidential data. As part of the recruitment process, the selected candidate might undergo a background check.
→ META: master thesis, exjobb, examensarbete, degree project 2027, computer vision, deep learning, perception, self-supervised learning, world models, 3D reconstruction, Gaussian splatting, JEPA, autonomous driving, Gothenburg, Göteborg, Sweden, paid master thesis, Zenseact, Volvo Cars
More about Zenseact
Our software makes a difference.
Using AI-based technology to create the ultimate driver support, we’re fighting to end car accidents and make roads safe for everyone. Every year, around 1,4 million people die in traffic while approximately 50 million people get injured. Many get disabled as a result of their injury. We can do better.
One purpose, one product.
We’re a software company dedicated to revolutionizing car safety. By designing the complete software stack for autonomous driving and advanced driver-assistance systems, we’re fighting to end car accidents and make roads safe for everyone. Zenseact was founded by Volvo Cars, and the teams are based in Gothenburg and Lund, Sweden and Munich in Germany. When we aim for zero accidents faster, we strive to speed up the transition to safe automation. This is essentially achieved by making cars updatable - like a computer or a phone. With regular software updates, a vehicle can be made safer long after its production. By accelerating improvement loops, shortening development cycles, and deploying high-capacity software quickly, we can make cars safer, and faster.
Culture with people at heart
To achieve our mission of saving lives and ending traffic accidents we must go where nobody has before. It requires us to venture into the unknown, pioneering new technology and pushing the frontier of autonomous driving. While there’s no denying our determination and expertise, we must stand united to succeed. By fostering a culture of support and enablement - a place of psychological safety where all of us can thrive - everything else will follow. We call this a people-at-heart culture. This culture means caring. It means the company cares about me, and we care about one another. It means sharing, so we give each other energy and have fun together. Our culture is also about belonging. It’s important to feel at home and that we can be ourselves at work. Finally, a people-at-heart culture means well-being. So, we enjoy the flexibility needed to be and do our best - at work and in life.
Zenseact works proactively to create a culture of diversity and inclusion, where individual differences are appreciated and respected. To drive innovation we see diversity as an asset, which means we value and respect differences in gender, race, ethnicity, religion or other belief, disability, sexual orientation or age, etc.
Interviews are held continuously, so we highly recommend that you submit your application at your earliest convenience.