# A Developer’s Guide — Integrating Verifiable AI Today

Hey ML folks,

So, you're into crypto but think integrating proofs into your PyTorch model is tough?

Check out VerifiableModule.

It’s basically a wrapper you can just plug into your current nn.Module. Instead of redoing your model, just wrap it like this:

from cysic import VerifiableModule

model = VerifiableModule(existing_pytorch_model)
output, proof = model(input_tensor)

This thing does all the heavy lifting:

1. Turns model operations into verifiable circuits.
2. Runs proof generation super fast using your GPU.
3. Gives you both the prediction and the proof, all set for validation.

Under the hood, it uses tons of CUDA threads for sumcheck proofs, and it uses high-speed memory for field arithmetic. It's set up to use memory and math units as efficiently as possible.

Right now, it works best with CNNs and models that aren't too big. They're working on support for LLMs. Keep an eye out for open-source toolkits, demos, and quick start guides coming soon.