Fix grammar

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Author: izzet

src/pages/research/projects/viper.mdx

@@ -112,7 +112,7 @@ During training, DNN models can be cached on multiple alternative locations (e.g
     - Creates a direct communication channel to transfer the model data
         - Direct GPU-to-GPU memory and host-to-host memory data transfer strategy
         - Implemented based on MPI library
-3. **A lightweight publish-subscribe notification module** to proactively informing consumers of model updates instead 
+3. **A lightweight publish-subscribe notification module** to proactively inform consumers of model updates instead 
 of passively periodic queries