Timely and safe AI acceleration for time-sensitive AI applications
ai
91% of all Companies rate safety and transparency as important or very important in the AI transformation
PWC Study
What we do
AI SAFETY FIRST
AI Processes are complex in that they have to account for completeness, correctness and coherence of the training and evaluation steps during the development of the AI models
Additionally it has to account for top level processes + data management so that the data being used is correct, complete and has no bias.
For this an AI needs computer hardware, just as our ideas need a brain which we provide
For example: Safety mechanisms must ensure the integrity of the hardware in order to prevent corruption of data, else it can lead to wrong decisions of the AI
R & D
Our technology builds on years of groundbreaking research at the Institute for Embedded Systems at the University of Siegen, which is very well known for its research in reliable real-time systems.
NEURAL NETWORKS
We provide an end-to-end solution for developing your AI applications with Tensorflow, PyTorch and ONNX.
NEURAL NETWORK ACCELERATOR
Our technology is powered by an FPGA which contains our safe AI inference technology providing guaranteed AI inference latency without any interference of other AI applications running at the same time on the accelerator.
PROCESS AUTOMATION
Good synchronization of manufacturing processes increases production and prevents downtime. However, to achieve good synchronization, the duration of processes and computer calculations must be as stable as possible. This is a particular challenge when using AI processes. With our guarantee of constant inference times, this is no longer a problem for our customers.
OUR PARTNER NETWORK
OUR TARGET MARKETS
Factory Automation
In the near future, AI will enable humans and robots to work together (CoLabs). Of paramount importance is the safety of humans, where any failure of AI must be prevented.
Smart Home Facility Equipment
AI in smart homes is already well established and is currently making its way into professional building equipment. We support this transition for safety-critical use cases such as automatic doors and intelligent emergency systems.
Patient monitoring and Robotic Surgery
AI helps surgeons operate more precisely than they ever imagined. However, time-critical operations require reliable AI that always provides help in a timely manner.
Autonomous Driving
Autonomous driving is inconceivable without AI. Here, it is a matter of necessity that AI must provide reliable and timely assistance at all times.
THE THEORY
A fundamental aspect of safety-critical systems is determinism. Meaning that the same inputs provide the same outputs, every time and always within a given deadline.
When we talk about AI, this is very challenging as the set of inputs is infinite. But the outputs are calculated by a neural network which is trained by a finite training dataset.
So, how can we deal with this?
We need assurance of AI itself (so called Trusted AI) along with AI hardware integrity and timeliness!
Trusted AI Trusted AI is concerned with the completeness, correctness and coherence of the training and evaluation steps. It covers all of the top-level safety processes plus data management. This is to make sure that the data being used is correct, complete and has no bias.
Trusted AI is concerned with the completeness, correctness and coherence of the training and evaluation steps. It covers all of the top-level safety processes plus data management. This is to make sure that the data being used is correct, complete and has no bias.
Integrity and Timeliness of AI-Hardware At the inference level, safety mechanisms ensure the integrity of the hardware in order to prevent corruption of data. Further, scheduled execution of AI inference assures timeliness of the results.
At the inference level, safety mechanisms ensure the integrity of the hardware in order to prevent corruption of data. Further, scheduled execution of AI inference assures timeliness of the results.
ADVISORS
Prof. Roman Obermaisser
Scientific Advisor
University of Siegen
Prof. Bernd Buxbaum
Technical Advisor
PMD Technologies AG
Prof. Martin Hill
Advisor Entrepreneurship
Hill GmbH
Prof. Ali Jennesari
Scientific Advisor
University of California, Berkeley
Frank Ermert
Founders Network
University of Siegen
Contact Us
We will get back to you soon