CANDiY AI begin by compressing large AI model and deploying them onto user devices using knowledge transfer techniques.
The size of the AI model is optimized based on the environment of he user device.
Once deployed, the basic AI model on the device accesses the user’s private and sensitive data for further learning and customization.
To maximize the performance of these customized AI models, CANDiY AI’s personalized federated learning algorithm is applied.
Next Gen Federated Learning
The algorithm allows for the sharing of AI knowledge among users with similar profiles while ensuring that all user data remains securely on their devies.
Throughout the process, user data remains on the device, and only the AI parameters trained on the device are shared, ensuring data privacy.
AI parameters are encrypted, ensuring that external parties cannot access the distribution or composition of the data within the device.
