• high cost
  High costs due to increased computing resources and network resources for handling huge amounts of data generated from IoT devices
• delay
  Difficulty in realizing real-time processing due to data processing delay
• privacy
  Privacy protection concerns due to information collection from customers and users

• STADLE™ collects only the parameters of the edge AI's local AI model, generates a global model that integrates them on the cloud AI side, and returns them to the edge AI side, making it possible to promote edge AI learning.
• Since raw data is not shared, conventional problems of pressure on computing and network resources and processing delays are resolved.
• Mitigate the risk of identity theft from data sharing and provide a high level of privacy protection

[Point 1] Multiple functions to accelerate AI learning

Features of TieSet STADLE™

[Point2] Overwhelming intelligence

Compared to AI estimation with edge alone, it can reach the optimal solution faster.

[Point 3] Application areas where the use of TieSet's STADLE™ is expected

TieSet's STADLE™ protects personal information, can be used in a distributed environment, and is capable of real-time processing. ) are expected to be used in the automotive field.

Core features of STADLE™

Distributed horizontal scaling with Kubernetes

Common federated learning platforms cannot scale horizontally and end up constructing a single global model (aggregation of cluster models) for all devices. However, the global model may not be suitable for some attributes. TieSet's STADLE™ allows you to flexibly form multiple semi-global models to improve model performance.