The challenge

The power of artificial intelligence (AI) is virtually limitless and is widely used, for example, to optimize manufacturing processes, medical diagnostics and autonomous mobility. AI now often still runs in the cloud. Devices collect data, for example with sensor technology, and then send it to the cloud for processing with intelligent algorithms.

With Edge AI, on the contrary, these algorithms run locally: on the device itself. Thus, there is no need for a continuous connection to the cloud and the constant sending of data back and forth. This offers great advantages for mission-critical and time-sensitive applications: it prevents latency/delays and increases reliability, cybersecurity and cost efficiency. So the promise of Edge AI is great. However, moving from the cloud to the edge is technologically challenging, and making the right choices requires complex trade-offs of performance, cost, power consumption and ease of integration. Sioux helps customers through this transition.

The solution

AI models are typically developed for high-performance high-tech systems with virtually unlimited resources. But how do you ensure that algorithms work locally and perform maximally within the constraints of the real world, for example in terms of computing power and power consumption? First of all, it requires systems thinking and combining specialized knowledge about artificial intelligence, electronics and embedded software. In addition, it requires, among other things, to connect seamlessly to the customer's specific information needs and technology.

Here lies the strength of Sioux. Our solutions in the field of Edge AI include adapting existing algorithms for local use, optimizing AI models for the best possible performance through to product development and maintenance. Thus, Sioux supports customers in the development and implementation of robust, affordable and scalable Edge AI that makes a real difference in the success of their applications.


Sioux supports customers with a systems approach to Edge AI. We provide the electronics selection and artificial intelligence modifications required, ensuring the best outcome in the combination of performance, cost and other requirements.

Edge AI services

Solution modeling
If a customer has not already done their own preliminary study, Sioux applies the method that best fits the customer problem, for example Model Predictive Control, Reinforcement Learning, Optimized Tensor Computing or Bayesian inference for low-power devices.
Technology selection
Sioux has deep knowledge and extensive experience in the use of GPUs, CPUs, FPGAs, TPUs and full platforms. Think Nvidia Jetson, Coral, Vitis, Utrascale+, ARM, or our own Sioux high-performance reference architecture.
Model implementation
Every hardware choice affects the performance of an algorithm. We make the selection of the right platform - for example, TinyML, Tensorflow Light or our proprietary libraries for Bayesian inference - within tight-knit multidisciplinary teams.
Life Time Support for Edge AI
For the purpose of maintenance and tutoring on data (MLops), we use AWS, Google Cloud, MS Azure, Kubernetes, MLFlow, Apache Airflow, Kafka, Faust and Kubeflow, among others. Our holistic process has an additional layer above that for deployment to the specific hardware.

Contact us here