Senior AI Inference & Optimization Infrastructure Engineer

DiDi; San Jose (CA); On-site; USD 170,000 - 351,000; Full time

Job summary

DiDi Autonomous Driving seeks a Senior/Sr. Staff AI Infrastructure Engineer, Inference & Optimization, to lead performance tuning and deployment of AI models across vehicle-side and cloud infrastructure.

You will design high-efficiency inference pipelines, build system-level stability frameworks, and optimize hardware execution for ultra-low latency in autonomous systems. You will collaborate with cross-functional teams to accelerate model iteration and ensure rock-solid, production-grade

Qualifications

• Master's degree or higher in a closely related technical field.
• 3+ years in high-performance computing, AI infrastructure, or embedded deployment.
• Strong skills in C++ and Python with CUDA/OpenMP experience.
• Familiar with TensorRT/ONNX Runtime and LLM inference frameworks.
• Knowledge of modern GPU architectures and memory bandwidth management.

Responsibilities

• Own deployment, optimization, and resource scheduling of vehicle-side AI models.
• Lead stability initiatives and root-cause analysis for performance bottlenecks.
• Architect and scale deployment environments for LLMs and foundational models.
• Track industry methodologies and implement advanced optimization toolchains.
• Establish profiling/telemetry using CUDA tools across GPU architectures.
• Collaborate with perception/prediction, cloud infra, and safety teams.

Skills

C++

Python

CUDA

OpenMP

Profiling tools

Education

Master's in Computer Science

Tools

TensorRT

ONNX Runtime

vLLM

SGLang

TensorRT-LLM

Job description

DiDi Autonomous Driving seeks a Senior/Sr. Staff AI Infrastructure Engineer, Inference & Optimization, to lead performance tuning and deployment of AI models across vehicle-side and cloud infrastructure.

You will design high-efficiency inference pipelines, build system-level stability frameworks, and optimize hardware execution for ultra-low latency in autonomous systems. You will collaborate with cross-functional teams to accelerate model iteration and ensure rock-solid, production-grade