Thoughts after the.NET AI presentation

On March 22, 2025, I gave a talk on.NET AI in Shenzhen, China, where I explained a bit about TorchSharp and ML.NET.


Event information:
[https://globalai.community/bootcamp/china-shenzhen/](https://globalai.community/bootcamp/china-shenzhen/)


Since PPT can only hold a little content, I created a separate Github warehouse to demonstrate cases and explain the detailed process.
[https://github.com/whuanle/20250322](https://github.com/whuanle/20250322)

I have a lot of readers in China who are also interested, in the Internet of Things and various fields, many developers also want to be able to achieve AI model training and reasoning through the.NET platform.


I think there are a couple of things that need to happen right now to really be able to train and deploy AI through the .NET platform.

* TorchSharp continues to synchronize some of Pytorch's network structures, for example.
* ML.NET is combined with TorchSharp, and the AI models trained by TorchSharp can be directly converted to.onnx format.


If I can do these two things, I can quickly implement more cases, use them in enterprise projects, and publicize and demonstrate them to readers.