Worked on the h2r/GHOST repository to advance robot perception by building a multi-robot depth completion and AI-driven perception feature. Developed C# scripts and integrated Unity’s Sentis library to process and fuse depth and color data from multiple cameras, resulting in a richer 3D environmental map for safer navigation and improved fleet collaboration. Incorporated ONNX-based depth estimation into the perception pipeline, enabling real-time AI-assisted depth processing from various viewpoints. Leveraged skills in 3D rendering, AI/ML, and computer vision to establish robust cross-robot data fusion, enhancing obstacle detection and scene understanding. All work emphasized reproducibility and clear code traceability.