AI at Embedded World 2020 with Phytec’s Dr. Jan Werth

Posted by – March 28, 2020

In this interview Dr. Jan Werth from the German hardware developer and manufacturer PHYTEC talks about artificial intelligence and embedded hardware. In the interview many topics are touched on and explained from the point of view of a data enthusiast and artificial intelligence expert. While many topics are touched upon, the interview focuses on artificial intelligence on the edge and embedded hardware.

If the interview has sparked your interest and you have further questions about a specific topic, you can contact Jan Werth ( or directly book an appointment with him ( ).

Further information about PHYTEC can be found here:

If you are interested in upcoming AI or embedded workshops, you can find them here:

The hardware presented in the Video (i.mx8MQ) can be found here:

Outline of the interview:

Introduction of Jan Werth.
What is happening in artificial intelligence in the embedded world ?
· The start of AI for embedded hardware
· Idea of edge processing
· Security from the edge
· Misconceptions
· Phytec strategy
· New AI-Chips and dedicated hardware is needed.
· Is the embedded industry behind in AI-chip solutions?
· You do not need dedicated AI chips ◦ More details on the AI-Chips ◦ There are other problems than autonomous driving ◦ Misconception that you always need super cluster for AI ◦ Longterm support with AI-Chips
· New chip from NXP with added AI acceleration ◦ Gyrfalcon / Nvidia jetson / Microsoft Movidius/ Google coral TPU
· Why Phytec • Presentation of Demos ◦ With Microsoft Azure (hand gesture recognition) using Azurs Iot Edge agent
· Training a model from scratch (facial detection and matching embeddings to celebrities)
· What takes how long in the AI demo. Bottleneks in the AI process? ◦ What belongs in a AI pipeline?
· Future of AI in embedded Hardware? ◦ Hype or not? ◦ The AI winters. ◦ Revolution in embedded hardware with AI-chips. ◦ Software vs hardware • Makes Edge AI more green?
· AI projects with Phytec? ◦ What are the different types of the embedded costumers using AI • How hard is it to get into AI for an embedded hardware/ software developer?
· How do you process your data correctly? • Tips for the management. ◦ What to expect form a AI project. ◦ What takes how long? • Open Source and AI. ◦ Open Source datasets. ◦ AI libraries • Who will be the winner – open source or propriety ? • How do you synch your data best? • What would be great to use AI on edge for?

Filmed at Embedded World 2020, watch all my Embedded World 2020 videos in this playlist here: