Machine Learning Applied Scientist - ADAS Online

Job description

We are building a high-performance ADAS Online team focused on advancing state-of-the-art machine learning and AI algorithms for scene understanding and environmental awareness. As a machine learning applied scientist, you will contribute to the development and improvement of models within the in-vehicle spatial awareness stack. Work hands-on across data, experimentation, and evaluation pipelines, and grow your ability to tackle increasingly complex engineering challenges in a fast-moving team. This spatial awareness stack will leverage modern transformer and end-to-end architectures to transform vehicle sensor data along with real-time 3D map data from the cloud into a 3D, semantically defined environment identifying all static and dynamic objects. What You’ll Do Develop a solid understanding of the team's physical AI roadmap and contribute ideas that improve algorithms within your area of ownership; Implement and improve ML / vision transformer models for 3D awareness; Produce instance detections and segmentation masks, instance tracking, and occupancy modelling; Contribute to multi-modal fusion pipelines (camera, LiDAR, RADAR), building and validating components against defined requirements; Apply Transformer-based and representation learning techniques within established frameworks, building confidence in large-scale model development; Run experiments and benchmarks with guidance, producing clear results and contributing to the team's evaluation practices; Participate in translating research into production, learning what makes ML solutions reliable and scalable; Actively seek feedback from senior engineers and apply it to continuously improve the quality of your work. What You'll Need 3+ years of experience in machine learning, vision transformers, diffusion, or computer vision; Expertise in modern deep learning architectures; Hands-on experience with PyTorch (or equivalent frameworks); Understanding of what is needed to build and deploy large-scale ML models; Strong ownership and bias toward execution; Data-driven mindset with rigorous evaluation practices; Ability to operate effectively in fast-paced environments; Clear communication of technical trade-offs; High standards for code quality and reproducibility; Comfortable navigating ambiguity in complex technical domains. Nice to Have Experience in autonomous systems or robotics perception; Publicati…