diff --git a/README.md b/README.md index d994be5..81144f2 100644 --- a/README.md +++ b/README.md @@ -61,6 +61,13 @@ Forget multi-million dollar on-prem infrastructure, get the same privacy guarant +
+ + + inference with and without equivariant encryption diagram + +
+ ## How Nesa Achieves Blind AI: Equivariant Encryption (EE) ## At Nesa, privacy is a critical objective. On our path toward universal private AI, we confronted a key challenge: **how can we perform inference on neural networks without exposing the underlying input and output data to external parties, while returning requests without high latency?** Traditional approaches, such as differential privacy, ZKML or homomorphic encryption (HE), while conceptually strong, fall short in practical deployments for complex neural architectures. These methods struggle to handle non-linear operations efficiently, often imposing substantial computational overhead that makes them infeasible to integrate into real-time or large-scale systems. diff --git a/docs/ee-light.png b/docs/ee-light.png index f891048..cdec716 100644 Binary files a/docs/ee-light.png and b/docs/ee-light.png differ diff --git a/docs/tokenizer-light.png b/docs/tokenizer-light.png index c04af20..6c892e0 100644 Binary files a/docs/tokenizer-light.png and b/docs/tokenizer-light.png differ diff --git a/docs/tokenizer.png b/docs/tokenizer.png index d34033b..58c517a 100644 Binary files a/docs/tokenizer.png and b/docs/tokenizer.png differ