Full-Stack AI Systems Development with Murali Akula - #563
Published:Mar 14, 2022 16:07
•1 min read
•Practical AI
Analysis
This article from Practical AI discusses the development of full-stack AI systems, focusing on the work of Murali Akula at Qualcomm. The conversation covers his role in leading the corporate research team, the unique definition of "full stack" at Qualcomm, and the challenges of deploying machine learning on resource-constrained devices like Snapdragon chips. The article highlights techniques for optimizing complex models for mobile devices and the process of transitioning research into real-world applications. It also mentions specific tools and developments such as DONNA for neural architecture search, X-Distill for self-supervised training, and the AI Model Efficiency Toolkit.
Key Takeaways
- •Qualcomm's approach to full-stack AI development, encompassing research, deployment, and optimization for Snapdragon chips.
- •The challenges and techniques involved in deploying complex machine learning models on resource-constrained mobile devices.
- •Specific tools and methods like DONNA, X-Distill, and the AI Model Efficiency Toolkit for improving AI model performance and efficiency.
Reference
“We explore the complexities that are unique to doing machine learning on resource constrained devices...”