Staff AI Application Engineer (Automotive SoC)

Job description

Job Summary

We are looking for an AI Application Engineer to support the enablement, optimization, and deployment of AI models on automotive-grade SoCs. In this role, you will work closely with internal compiler/runtime teams and external customers to bring AI models from training to optimized inference on embedded NPU/DSP platforms , with a strong focus on performance, accuracy, and system integration .

Key Responsibilities

AI Model Enablement & Optimization Enable and deploy AI models (e.g., BEV, object detection, segmentation, classification ) on Gen4/5 SoC platforms with CNNIP/DSP/NPU HWA. Perform model performance analysis (latency, throughput, multi-core scaling) and identify bottlenecks related to memory bandwidth, scheduling, or operator mapping.

Support model optimization workflows , including

Post-Training Quantization (PTQ) Quantization-Aware Training (QAT) collaboration Operator fusion, graph optimization, and execution partitioning Analyze accuracy degradation caused by quantization or operator limitations and propose mitigation strategies. Embedded AI Inference & System Integration Integrate AI models into embedded runtime environments (Linux / QNX).

Debug issues related to: CNNIP/DSP/NPU offloading Memory allocation / IPMMU Data transfer overhead and multi-core synchronization Validate AI workloads on target boards and simulators (SIL / HIL). Toolchain & Model Workflow Support Work with AI compiler and runtime toolchains (e.g., ONNX-based workflows, hybrid compiler, MWMX ).

Support ONNX model handling, including

Graph inspection and modification Model segmentation and execution control Quantized (QDQ) ONNX models Develop or maintain internal tools and scripts to improve model validation, benchmarking, and customer workflows. Customer & Cross-Team Collaboration Act as a technical interface between customers, internal development teams, and field application engineers. Support customer evaluations, PoCs, and demos on automotive AI platforms.

Provide technical guidance, documentation, and best practices for AI model deployment. Contribute to weekly technical reports, issue tracking, and release validation activities .

What We Offer

Opportunity to work on state-of-the-art automotive AI platforms . Hands-on involvement across the entire AI lifecycle , from model to silicon. Close collaboration with compiler, runtime, and hardware architecture teams. Exposure to real customer deployments and production constraints.

Qualifications

Required Qualifications

Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, Embedded Systems, or have experience in embedded systems . Solid understanding of deep learning fundamentals and inference pipelines. Hands-on experience with AI frameworks such as PyTorch, ONNX, or ONNX Runtime. Strong programming skills in Python ; working knowledge of C/C++ is a plus. Familiarity with embedded systems and debugging tools. Ability to analyze performance using metrics such as latency, throughput, and hardware utilization.

Good communication skills in a multi-cultural, cross-functional environment. Preferred / Optional Qualifications 1–3 years of experience in embedded systems or AI-related development . Experience with AI model training, fine-tuning, or evaluation , especially for: Computer vision models (Detection / Segmentation / BEV) Automotive or robotics use cases Practical experience with AI inference optimization on embedded hardware (NPU, DSP, GPU, or CPU).

Familiarity with quantization techniques (INT8, calibration methods, QDQ models). Experience with automotive SoCs or safety-related software environments (QNX is a plus). Understanding of memory hierarchy, DMA, and multi-core scheduling in SoC architectures.

Nice to Have

Experience supporting customers or acting in a technical support / application engineering role . Knowledge of automotive AI standards or ADAS perception pipelines. Experience contributing to internal tools, scripts, or documentation. Ability to read and debug ONNX graphs or intermediate representations.