Research project
Machine-Learned Jet Time and Trackless Jet Vertexing
Reconstructing particle interactions with timing and learning.
What can precise timing reveal about particle interactions when conventional tracking is insufficient?
Approach
Combine collider timing data and deep learning, then use time-based reconstruction to recover displaced vertices.
My contribution
- Developed the data-generation, normalization, and parameter-scan workflow in TensorFlow and C++.
- Built a time-based reconstruction model for displaced vertices.
Results
4×Improvement in jet-time precision
The model achieves fourfold better jet-time precision than prior methods and reconstructs displaced vertices in settings where conventional tracking is insufficient.