Your Role
Key responsibilities are as follows:
- Develop end-2-end AI tools, models or software to run quantized neural network models .
- Design and develop end-to-end test strategies and frameworks for complex software systems.
- Create and maintain automated test suites.
- Develop debugging/profiling tools and Qualcomm SDK.
- Participate in code reviews and contribute to improving test coverage.
- Collaborate with cross-functional teams to understand requirements and design test approaches.
- Analyze test results and provide detailed feedback to development team.
- Conduct experiments to reproduce and optimize accuracies and performance of models.
- Related papers reading and summarization and presentation.
About You
Ideal candidates will have:
- Currently pursuing or recently completed a degree in Computer Science, Artificial Intelligence, EE, or a related field.
- Experience in large language model (LLM), vision-language model (VLM), and large vision model (LVM), including diffusers, multi-modality, CNNs, RNN/LSTMs, Transformer, and others.
- Experience in Model quantization and compression, deployment in smartphones or other edge devices.
- Strong mathematical skills - good foundations of linear algebra, matrix, differential algebra, statistics.
- Have strong skill in C++ and Python programming.
- Experience in machine learning/deep learning algorithms and architectures, including CNNs, RNN/LSTMs, Transformer, LVM, LLM.
- Hands-on experience with ML frameworks, such as TensorFlow, PyTorch and Onnx Runtime.
- Have knowledge in on-device AI models deployment is a plus.
- Have knowledge in llama.cpp or ExecuTorch is a plus.
- Have knowledge in Qualcomm SNPE/QNN SDK is a plus.
Benefits
Qualcomm offers competitive compensation, health benefits, and opportunities for professional growth.
Training & Development
Access to continuous learning programs, mentorship, and development opportunities to enhance your skills.
Career Progression
Potential for growth into full-time roles within Qualcomm's AI research and development teams.