diff --git a/docs/gsoc2025/ideas.md b/docs/gsoc2025/ideas.md
index e62101f3..e2dedeff 100644
--- a/docs/gsoc2025/ideas.md
+++ b/docs/gsoc2025/ideas.md
@@ -17,6 +17,15 @@ LDTK exports raw level data in JSON format, which can be further parsed by game
 
 This project aims to develop a full-featured python library for importing LDTK levels, with advanced support of [Auto Tiles](https://ldtk.io/wp-content/uploads/2020/11/autoLayer-demo2.gif) for games with random map generation. The library should be written in pure python, compatible with pocketpy and cpython. If successful, it will be published on [PyPI](https://pypi.org/) and benefit all python game developers.
 
+### Development `cTensor` library for machine learning
+
++ Difficulty Level: 3/5 (Medium)
++ Skill: C; Further Mathematics
++ Project Length: Medium
+
+pocketpy is planning to provide a tensor library `cTensor` for users who want to integrate neural networks into their applications. `cTensor` implements automatic differentiation and dynamic compute graph. It allows users to train and deploy neural networks on client-side devices like mobile phones and microcontrollers (e.g. ESP32-C3). We have a early prototype located at [pocketpy/cTensor](https://github.com/pocketpy/cTensor).
+In this project, students will help develop and test the `cTensor` library, which is written in C11. We expect students to have a good understanding of further mathematics and C programming.
+
 ### VSCode plugin for debugging pocketpy applications
 
 + Difficulty Level: 3/5 (Medium)