DiDi Labs
Software Engineer – Map Fusion & Planning
About this role
DiDi Autonomous Driving seeks a Software Engineer to develop map fusion and motion planning systems for Level 4 autonomous vehicles. You'll design scalable infrastructure integrating HD maps, real-time sensor perception, and trajectory generation while optimizing deep learning models from training through efficient C++ runtime deployment.
What you'll do
- Architect data pipelines and APIs for map fusion, real-time vectorization, and motion planning modules
- Integrate offline HD maps with online perception to create unified local environment models
- Implement DETR-style, query-based vector decoding networks in bird's-eye-view for map element generation
- Design and validate motion planning algorithms with feedback loops between mapped features and trajectory optimization
- Own end-to-end deployment of deep learning models from Python training through ONNX to C++ runtime execution
- Develop real-time map anomaly detection and safety validation systems for planning reliability
What they're looking for
- C++ (performance-critical systems)
- Python (model training and development)
- Deep learning frameworks (PyTorch, TensorFlow)
- ONNX optimization and model deployment
- Motion planning and trajectory optimization algorithms
- DETR and transformer-based architectures
- Real-time systems and multi-threaded programming
- Autonomous driving and robotics fundamentals
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
DiDi Labs
DiDi Labs develops autonomous driving technology for Level 4 vehicles, focusing on motion planning, behavioral decision-making, map fusion, and robotic simulation systems. The company is hiring software engineers and researchers to build algorithms, trajectory optimization systems, and cloud-based simulation infrastructure that enable safe and intelligent autonomous vehicle navigation.
View all jobs at DiDi LabsLikely interview questions
- Describe your experience integrating offline HD maps with real-time sensor perception data. What challenges did you face?
- How would you design the data pipeline architecture to ensure low-latency communication between map fusion, perception, and planning modules?