In early July, I was honored to have a chat with Megha Daga, Senior Director of Product Management and AI/ML lead for the Internet of Things (IoT) at Qualcomm Technologies, Inc. . As a critical player in AI enablement across the IoT group at Qualcomm, Megha has been crucial in the development of cutting-edge AI solutions used around the world.
We dove right into what Qualcomm has been up to as it continues to advance IoT through the different core offerings, partnerships, and cutting-edge solutions that Qualcomm offers. To set the tone of our conversation, we discussed Edge AI.
Edge AI is essentially intelligence moving to the data generator, according to Megha. Along with getting data faster, a host of other factors impact Edge AI, including privacy, cost, latency, reliability, and bandwidth. For businesses or enterprises, the simplicity of the technology revolves around business intelligence occurring on the device or close to the device itself to enable IoT.
Qualcomm provides a portfolio of hardware technology, but even more exciting is their advancements in software design and embedded processing innovation. The company understands how heterogenous computing makes AI possible and is pushing the envelope to remain competitive in AI and IoT. Some of the stronger vertical markets and industries that Qualcomm is targeting include retail, logistics, energy, utilities, industrial, and robotics.
To further advance into AI, Qualcomm launched the Vision AI Development Kit. This Azure IoT Starter kit is a vision AI developer kit for running artificial intelligence models on devices at the intelligent edge. With Edge AI, data is generated and pushed to the cloud. Legacy devices such as retail payment terminals and other industry specific devices are being digitized and modernized. Hardware or devices can be connected to a box, i.e., edge gateway. Megha shared that Qualcomm is taking metadata and compute to the box, implementing further compute as needed, then sending only the required data back to the cloud. The traditionally “dumb” environment is becoming more intelligent and bringing efficiencies to businesses and operations. Another Qualcomm AI example outside of retail is in logistics, more specifically warehouse operations. Robotics and drones may be used for picking and dropping, reducing overall payloads, and therefore reducing costs.
Edge AI and IoT are coming together to minimize compute to the cloud, as the overall costs of sending massive data to the cloud is becoming more cost prohibitive, and a concern for larger enterprises. The issues of privacy, latency, and connectivity again remain important factors. Privacy not only affects consumers, it also impacts businesses and their customers’ experiences. As for latency, think of delivery robots on the street, providing sub-millisecond intelligence and information to enable operations and efficiency so consumers can get food, packages, products delivered (similar to same day delivery).
Regarding connectivity, especially for operations in remote locations (construction, agriculture), having on-device or near device data intelligence can be critical. Examples Megha mentioned included drones connected to a gateway to enable crop intelligence, construction management, and mining operations.
Qualcomm’s portfolio continues to evolve to support AI and Edge AI, with a stronger focus on software. Their hardware and chipsets will continue to be their foundation, as they grow their partnerships with Original Equipment Manufacturers (OEMs). Qualcomm is leading in the areas of enabling AI on traditional CPUs/DPUs or AI on SDKs. Another cutting-edge development includes AI on embedded processing (low power, high performance).
According to Megha, a few exciting AI areas that Qualcomm has been innovating around includes drone robots, and camera technology. Taking regular cameras for example and making them intelligent, using technologies such as machine vision and AI running on heterogenous computing to completely disrupt its capabilities. Megha shared that Qualcomm is using hardware accelerators for neural network workloads. Furthermore, AMR devices (autonomous mobile robots, i.e., Bosch devices) is an area where Qualcomm is developing chipsets and reference designs to further advance delivery. For example, they recently launched the RB6, a high-end chipset with an accelerator card allowing the robot to greatly improve throughput (i.e., delivery robots).
As far as software goes, Qualcomm is investing and innovating to provide seamless software across the Qualcomm AI stack. Qualcomm is providing unification for developer building and changes, using Qualcomm Intelligence multimedia SDKs providing authentication and simplification for development and deployment, across multiple verticals. Developers and software tools remain a top priority for Qualcomm. Qualcomm realizes the end-customer (businesses and government) require and need end-to-end solutions and thus continues to build out its IoT partner portfolio (vendors, integrators, industry focused providers) focused on software/applications, platforms, and other solutions
I’ll end with a great use case example shared by Megha. The Qualcomm AI Engine runs ML models in IoT devices, such as a security camera that recognizes a family member and activates a smart lock to allow entry. Or an office building that allows employees onto an elevator based on a touchpad. This context showcases the importance of how Qualcomm is advancing AI and IoT to prepare tomorrow’s businesses and cities.
For more reading, please check out, “Qualcomm Advances Development of Smarter and Safer Autonomous Robots for Logistics, Industry 4.0, and Urban Aerial Mobility with Next-Generation 5G and AI Robotics Solutions”
Written by Stephanie Atkinson, CEO of Compass Intelligence
Covering hot topics in the industry, new research, trends, and event coverage.