End to end Real-time Interactive Edge AI analytics platform for Industry and Automotive

Projektinformationen

sa.engine

ID Finanzhilfevereinbarung: 190115974

DOI

10.3030/190115974

Projekt abgeschlossen

EK-Unterschriftsdatum 21 November 2022

Startdatum 1 Dezember 2022

Enddatum 30 November 2024

Finanziert unter

• The European Innovation Council (EIC)

Gesamtkosten

€ 3 601 250,00

EU-Beitrag

€ 2 499 999,00

Koordiniert durch

STREAM ANALYSE SWEDEN AB
Sweden

Periodic Reporting for period 2 - sa.engine (End to end Real-time Interactive Edge AI analytics platform for Industry and Automotive)

Berichtszeitraum: 2023-12-01 bis 2024-11-30

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

Stream Analyze operates in the dynamic field of edge analytics, with a focus on delivering an advanced edge AI platform to the manufacturing and automotive industries. Our overall objective is to enable organizations to leverage the power of edge analytics and AI by providing an end-to-end low-code platform that empowers data scientists, engineers, and domain experts, even those with limited coding skills. We address the increasing challenge of handling and analyzing the vast amount of data generated in real-time at edge devices. Sending all this data to the cloud for analysis is impractical and unsustainable, leading to a growing trend of performing analytics and AI at the edge, where the data is produced. Stream Analyze provides a groundbreaking solution to this problem.

By developing and refining our cutting-edge platform, based on extensive academic research from Uppsala and Stanford Universities, we are working towards becoming the preferred platform for large manufacturing and automotive companies seeking to bring efficient edge AI solutions to market. Our platform’s impacts are significant, as we anticipate facilitating faster, more efficient, and scalable development and implementations of edge AI solutions, ultimately contributing substantial value and cost savings to our customers.

Through our platform, we strive to address identified problems and needs in the manufacturing and automotive sectors. By enabling real-time AI and analytics models on massive fleets of edge devices and microcontrollers/MCUs, we empower our customers to extract valuable insights, optimize operations, enable servitization, and drive innovation. With Stream Analyze, organizations can capitalize on the full potential of edge computing and AI to enhance efficiency, productivity, and competitiveness.

Overall, our project sets the stage for a transformative journey in edge analytics, providing a robust foundation for organizations to navigate the challenges and seize the opportunities presented by the rapidly evolving digital landscape. More specifically the work involved include how to widen the target range of hardware to a greater number of platforms, and a methodology to quickly add new targets as they emerge on the market. The project also adds plug-in technology for common development tools, such as Microsoft's WS Code so developers can continue working with the tools they know and love, yet with new functionality.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

We made significant progress during the project. Work has been carried out in line with our work plan; to develop, improve and refine several areas of our technology and product offering. We have reached a maturity level where after completion can sell production-grade licenses to our customers also for large-scale deployments. The results include new possibilities for developers to access our technology by creating plug-ins to known IDEs and other code authoring environments, support for a significantly larger number of target platforms (hardware) used by the industry today, a new way of authoring functions and models using common SQL langugage experessions, and creating functions and libraries allowing developers with less or even no prior experience to create and deploy Ai models and visualisations tools for realtime streaming data.


We have spent significant time validating our offering with customers via customer projects and pilots targeting a number of real-world use-cases. The feedback we have received during the project verifies that we've progressed significantly since the project started and today our technology makes up a strong foundation for the continued development of the company. In further relation to the technology, the IPR has been safe-guarded, with 3 patents awarded and 14 pending covering important technical components and break-throughs, safe-guarding our key technologies on an increasingly fierce market within distrubuted edge computation and analysis.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

Our technology and the product offering are at the state-of-the-art for distributed edge computing. We have a core technology based on an ultra-small and efficient engine for streaming data, enabling work with real time data streaming on very resource-constrained devices. Our memory footprint for the engine was about 100 kb when we started the project. This is orders of magnitude smaller than comparable streaming engines. We have been able to further reduce the footprint of our engine down to less than 20 kb for the smallest instances available. This is an extraordinary achievement beyond even our own expectations.

We have also been able to improve the performance of our engine and the AI models running in the engine. Also in this area, we have achieved extraordinary results. We have reached a performance improvement factor up to 40x for numerical expressions in queries and models. In some cases, we have been able to improve the performance by a factor 1000 (thousand), dealing with streamed image information and large volumes of data.

The performance optimizations are due to an improved muti-step optimization process involving clever usage of a database query optimizer, a computer algebra, and a just in time compiler.

Stream Analyze logo