Summary
Kiran Kokilepersaud is a PhD candidate in Electrical and Computer Engineering at Georgia Tech specializing in foundational and applied self-supervised learning (SSL), with nine years of research and engineering experience. He develops theory-driven SSL methods—leveraging information theory to balance mutual information and representation dimensionality—and practical regularizers like HEX that improve local representational spread across multiple SSL algorithms. His work has been applied to medical imaging, autonomous driving, seismology, and multimodal clinical datasets, including creating a longitudinal ophthalmology dataset for biomarker detection. Kiran has industry experience probing model representations and building multimodal fusion at Apple and has a background spanning robotics simulation, cybersecurity research, and ASIC verification, reflecting broad systems and signal-processing fluency. He combines deep theoretical insight with hands-on data and tooling work, often exploiting domain-specific noise patterns (e.g., fisheye distortions) to improve robustness in real-world sensing.
9 years of coding experience
Doctor of Philosophy - PhD, Electrical Engineering, Doctor of Philosophy - PhD, Electrical Engineering at Georgia Institute of Technology
The University of Maryland, College Park