Senior AI Model Enablement Engineer – On-Device ML Tools

QUALCOMM, Inc.; San Diego (CA); On-site; USD 141,000 - 211,000; Full time

Benefits offered by this job

Benefits package

RSU grants

Annual bonus

Job summary

Qualcomm Technologies, Inc. in San Diego seeks an Engineer/Senior Engineer for the AI Stack SDK Software team to bring up and optimize modern AI models on Snapdragon platforms, focusing on on-device execution, quantization, and calibration across CPU/GPU/accelerator backends.

You will collaborate with AI Research, hardware, and software teams to bridge model architectures with backend frameworks, write high-quality Python and C/C++ code, and validate models for latency, power, and accuracy

Qualifications

• Strong programming skills in Python and/or C/C++.
• Foundational understanding of AI/ML concepts and model inference workflows.
• Experience with software engineering fundamentals: data structures, algorithms, debugging, testing, and source control.

Responsibilities

• Bring up and enable new language, vision, vision-language, and multimodal models on Snapdragon platforms.
• Study model implementations and graph patterns to optimize operators and data flow.
• Implement model re-authoring, graph transformations, and backend-specific adaptations.
• Develop reusable model support integrated into Qualcomm’s model-enablement frameworks.
• Convert, optimize, and export models from PyTorch and ONNX for compilation and execution.

Skills

Python

C/C++

Debugging

Education

Bachelor’s degree in CS/Engineering

Master’s degree (preferred)

Tools

Git

CMake

Linux

Job description

Qualcomm Technologies, Inc. in San Diego seeks an Engineer/Senior Engineer for the AI Stack SDK Software team to bring up and optimize modern AI models on Snapdragon platforms, focusing on on-device execution, quantization, and calibration across CPU/GPU/accelerator backends.

You will collaborate with AI Research, hardware, and software teams to bridge model architectures with backend frameworks, write high-quality Python and C/C++ code, and validate models for latency, power, and accuracy