Mr. Kunvar Chokshi

Edge AI in Embedded Systems – Bringing Intelligence Closer to the Real World

Abstract:

Edge AI is transforming embedded systems into intelligent, autonomous platforms capable of performing real-time decision-making without reliance on cloud infrastructure. This session explores the integration of artificial intelligence with resource-constrained embedded devices, highlighting advancements in low-power microcontrollers, specialized SoCs, and neural accelerators.

The discussion will focus on the practical challenges and design trade-offs involved in deploying machine learning models at the edge, including considerations for latency, energy efficiency, and model accuracy. Attendees will gain exposure to emerging toolchains such as TensorFlow Lite Micro and TVM, as well as deployment strategies used across industries like healthcare, automotive, smart manufacturing, and agriculture.

Real-world examples will demonstrate how edge devices are enabling smarter, faster, and more secure systems. The talk will also delve into future directions such as TinyML, federated learning on-device, and over-the-air model updates, illustrating the growing potential of embedded intelligence in decentralized and data-sensitive environments.

Profile:

Kunvar Chokshi is a distinguished Embedded Software Engineer with over 10 years of experience at industry-leading companies such as NVIDIA, Apple, and Tesla. Currently, he holds the position of Staff Software Engineer at Tesla in the USA. He earned his Master's degree in Electrical and Computer Engineering from the esteemed University of Maryland, College Park. His rigorous academic background, coupled with a decade of industrial expertise, has equipped him with a comprehensive skill set in software, IoT, robotics, embedded systems, and operating systems. Kunvar's profound understanding of complex systems and his proven track record of developing groundbreaking technology have enabled him to contribute to innovative projects that continually push the boundaries of possibility.