We held our next tinyML Talks webcast. Andrew Kostic from Qualcomm Technologies and Ehsan Saboori from Deeplite presented The 1st IN-PERSON Meetup in Toronto on December 16, 2022.
Welcome back, tinyML Toronto community, to our first in-person event! On December 16th, we will get together for a morning of discussion, presentation and collaboration at Qualcomm AI’s office in Markham, Ontario. There will be 2 exciting and informative presentations from leading Toronto AI experts at Qualcomm and Deeplite. There will be light snacks and refreshments provided by our host and attendees are encouraged to network before/after the event. All guests are required to register beforehand. SPACE IS LIMITED SO RSVP ASAP!
“An overview of the Snapdragon 8 Gen 2 Sensing Hub”
Speaker: Andrew Kostic, Sr. Director of Product Management, Qualcomm Technologies, Inc.
At the heart of the Snapdragon 8 mobile platform is the Qualcomm AI Engine. With the Snapdragon 8 Gen 2, we are updating the generation Sensing Hub with a dedicated dual-core AI Accelerator, for low power and always-on AI processing.
The Snapdragon 8 Gen 2 Sensing Hub with AI Accelerators offers double the performance of the previous generation, allowing for:
- Improved always-on camera
- Better AI-noise suppression, for better calls wherever you are, including your hybrid work
- New low power AI use cases developers are still thinking of thanks to its improved programmability and additional memory
In this talk, we’ll highlight some of the capabilities of the 4th gen Sensing Hub, shown at the recent Snapdragon Summit 2022 | Snapdragon Tech Event (qualcomm.com)
Andrew Kostic, Sr. Director of Product Management, Qualcomm Technologies, Inc.
Product management lead for Low Power AI along with Audio, and Voice technologies, across Mobile, Compute, XR, Automotive, and IoT. He has a background in computer engineering.
“Expanding possibilities for Edge AI with compressed & quantized CNN models”
Speaker: Ehsan Saboori, CTO and co-founder of Deeplite
The emergence of Deep Neural Networks (DNNs) on embedded and low-end devices holds tremendous potential to expand the adoption of AI technologies to wider audiences. However, making DNNs applicable for inference on such devices using techniques such as quantization and model compression, while maintaining model accuracy, remains a challenge for production. Ehsan will be presenting results from a novel inference engine (DeepliteRT) and compression framework (Deeplite Neutrino), developed by Deeplite Inc. He will discuss how Deeplite Neutrino automatically compresses and quantizes CNN models, and give real-world case studies of DeepliteRT, a first-of-its kind engine to run PyTorch deep learning models at real 2bit and 1bit precision.
Ehsan is CTO and co-founder of Deeplite. He obtained his BSc in artificial intelligence and completed his PhD in computer science at Concordia university in 2016. He has been a member of ACM and IEEE associations and published several patents, peer-reviewed conferences and journal papers. With several years of experience working in different companies such as Microsoft, SAP and Morgan Stanley he cofounded Deeplite where he assessed emerging challenges for deep learning and formed the technology vision that became the core of Deeplite.
=========================
Download presentation slides
Feel free to ask your questions on this thread and keep the conversation going!