Snapdragon and Qualcomm branded products are products of
Qualcomm Technologies, Inc. and/or its subsidiaries.
You might have seen a few headlines lately about AI, most notably generative AI. While most people tend to think that AI requires powerful, cloud-based servers, today’s edge devices, like mobile phones, are more than capable of running inference and training of large and complex models at the edge.
Qualcomm has been researching, developing, and deploying AI solutions to optimize AI for the edge for over 15 years. And with our technology, developers can run complex models optimized for size and power efficiency across devices from mobile phones to automobiles. We have so much information on this, that we created a course to break it down for you.
Fundamentals of Qualcomm AI is our new five-chapter course, in which we share what Qualcomm Technologies offers for AI at the edge. The course is for software developers, architects, test engineers, and product managers looking to build great products based on Qualcomm Technologies AI solutions.
In the first four chapters, Vinesh Sukumar, Senior Director, Product Management at Qualcomm Technologies, will take you on a journey of our AI solutions, starting from our initial AI research project in 2007. You’ll see how our AI hardware has evolved into the Qualcomm AI Engine, found in many Qualcomm and Snapdragon platforms that target a range of verticals, including mobile, automotive, compute, IoT, robotics, audio, and XR.
On the AI software side, you’ll learn about the Qualcomm AI Stack. This is our unified AI software portfolio to optimize and deploy AI models quickly with support for AI frameworks and runtimes, developer libraries, system software, and popular operating systems.
Figure 1 – Vinesh Sukumar explorers Qualcomm Technologies AI solutions
Figure 2 – Parham Noorzad dives deeper into AIMET and the AIMET Model Zoo.Parham Noorzad, Senior Product Manager at Qualcomm Technologies, joins in the last chapter to provide a deeper dive into AIMET. AIMET is significant for optimizing large ML models to run on constrained mobile devices with small, thin form factors and limited storage and battery capacity. You’ll also find out how the AIMET Model Zoo provides a repository of popular, open-source ML models optimized with AIMET that developers and researchers can use as a starting point in their work.
Ready to Learn More About Qualcomm Technologies AI Solutions?
Sign up for Fundamentals of Qualcomm AI – it’s free through the end of October. Then grab a coffee and settle in. The course is listed as two hours, but some of our students have taken it in as little as one hour.
Once you have the fundamentals down, here are some additional AI learning resources for you:
- Qualcomm AI Stack: Ecosystem of Qualcomm Technologies frameworks and tools for optimizing and running ML models on Snapdragon SoCs.
Some of the core components within the Qualcomm AI Stack include:- Qualcomm Neural Processing SDK: Tools for model conversion and runtime inference.
- Qualcomm AI Engine direct: Lower-level, unified APIs for AI development. Target a specific accelerator, or delegate workloads from TensorFlow Lite or the ONNX runtime to directly access our Qualcomm Hexagon processor.
- AIMET: Qualcomm Technologies toolset for quantizing and compressing ML models.
- AIMET Model Zoo: Repo of popular open-source ML models optimized by AIMET.
Snapdragon and Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries
AIMET and AIMET Model Zoo are products of Qualcomm Innovation Center, Inc.
