Edge AI promises to transform healthcare, industrial automation, consumer electronics and agriculture. Yet deploying it on embedded devices remains difficult. Models must meet demanding accuracy requirements within tight compute, memory and power limits, and developers need better ways to adapt them to their products.
In this session, we’ll explore DeepGate’s work on both challenges. On the R&D side, we’ll show how advances in compilers and model architecture make models more efficient, how developing new model operations and compiler support together can push those gains further, and where we see the greatest opportunities for future progress.
We’ll then turn to the platform: how developers can use AI coding agents and DeepGate’s tooling to adapt our efficient models to their own data, optimise them for their hardware and deploy them – shortening the path from idea to working product.