Jobs at DiDi

Research and implement motion and behavioral planning algorithms for autonomous vehicles, integrating geometry and velocity planning, modeling multi-agent interactions, formulating optimization-based cost functions, and validating performance in simulation and real-world data while collaborating with Perception, Prediction, and Control teams.

Design and implement map fusion, online BEV vectorization, and motion planning systems for autonomous vehicles. Build scalable C++/Python infrastructure, deploy deep-learning mapping models (ONNX→C++/TensorRT), optimize real-time planner latency and safety, and create anomaly detection for map/scene changes.

Lead system safety strategy and release readiness for unmanned autonomous driving. Define safety frameworks, release policies, V&V plans, safety metrics, incident analysis, and cross-functional collaboration to enable safe, scalable unmanned operations.

Lead architecture and implementation of low-latency IPC middleware, compiler/toolchain infrastructure, and deterministic simulation platforms. Own build systems, distributed caching, crash triage tooling, and performance/memory optimization across embedded autonomy stacks. Provide technical leadership, mentoring, and cross-team standards to improve developer velocity and system reliability.

Design and implement core motion planning and decision-making systems for autonomous vehicles, spanning high-level behavior (lane changes, merges) to low-level trajectory optimization. Develop geometry and velocity systems, model agent interactions, design trajectory costs, analyze and debug performance, and collaborate with Prediction, Perception, and Control teams to produce safe, comfortable, and efficient trajectories.

Design, implement, and maintain production-grade C++ planning software for trajectory candidate generation and scoring. Apply probabilistic and machine learning methods to evaluate trajectory risk and maneuver probabilities. Profile and optimize low-latency, memory-efficient planning on embedded platforms. Collaborate with Perception, Prediction, Motion Control, and Safety teams to validate planning behavior and metrics.

Lead company-wide AI-native transformation by designing and implementing a unified LLM/AI platform. Embed with autonomous driving teams to operationalize LLMs, optimize distributed training and inference, improve GPU utilization, and build resilient MLOps/LLMOps pipelines while setting technical roadmap and driving cross-functional adoption.

Lead deployment, optimization, and resource scheduling of vehicle and cloud AI models. Design low-latency inference pipelines, build system stability and profiling frameworks, optimize GPU execution, and scale LLM serving for simulation and validation. Collaborate with perception, cloud, and safety teams to enable reliable, production-grade autonomous deployment.