On Device AI - Deep Learning on Mobile Devices

sravya yellapragada (~sravya94)


Recent technological advancements in the field on Artificial Intelligence has lead to the development of wide range of mobile applications. Increased compute power, more efficient hardware and robust software, as well as an explosion in sensor data are fueling machine learning and moving actionable data and intelligence towards edge devices. Research and development of deep learning on mobile and embedded devices has attracted much attention. There is a wide improvement in performance of mobile hardware accelerators with every new generation of SoC. This in turn increases the capabilities of mobile deep learning frameworks to run complex AI models on device. On device AI helps in reducing latency, saving communication bandwidth, reduction of computing resource cost, quick response time and improving users data privacy. As AI makes devices, including smartphones and automobiles, more intelligent, mobile is becoming the key platform for enhancing all aspects of our lives, having an impact now and in the future.

In this talk we will cover the following aspects:

  • Different hardware accelerators for mobile deep learning like GPU, TPU, CPU etc in industry leader chipsets and their computing capabilities
  • Leveraging existing mobile deep learning frameworks like Tensorflow lite, Caffe2, Android Neural Networks API, Qualcomm Neural processing engine SDK etc to build efficient AI solutions
  • Short demo depicting the execution pipeline of deep neural net model on Android device
  • Optimization techniques like quantization, pruning, compression etc
  • Benchmarking for measuring the AI performance of mobile devices
  • Challenges and future scope


Basic knowledge on machine learning terms and techniques. Familiarity with any machine learning framework like pytorch, tensorflow etc

Speaker Info:

Hi, I am Sravya and I work in Qualcomm India Private Limited as a Software Engineer. I work on next generation software products based on AI. My responsibilities include developing powerful and efficient code for AI and machine learning apps to run smoother and faster on Qualcomm chipsets.

Speaker Links:

LinkedIn - https://www.linkedin.com/in/sravya-yellapragada-bb6684130

Github - https://github.com/sravyaysk

Twitter - https://twitter.com/sravyaysk

Section: Data Science, Machine Learning and AI
Type: Talks
Target Audience: Intermediate
Last Updated: