Bachelor thesis:

• In-depth understanding of gradient-boosting decision trees and the features of the LightGBM Framework (https://github.com/microsoft/LightGBM)
• Collection and review of available tools to port models to Arduino hardware and the evaluation concerning feature support and memory consumption

Master's thesis (or motivated bachelor thesis):

• Derivation of tools by identifying gaps within the existing tools - mainly concerning quantization (open for suggestions).

Good understanding of GBDT, and for a Master's thesis, basic C/C++ knowledge will help you.