My academic research across network measurement, agentic network experimentation, and computer vision.
The lab published a user-space DualPI2 module for Mahimahi in ACM SIGCOMM CCR.
I second-authored the DualPI2 paper, validated Mahimahi's fidelity against the Linux kernel with time-series and nonparametric analysis, and built concurrent experiments with under 5% cross-run variance.
The lab studies computer vision and 3D reconstruction.
I built a 3D Gaussian Splatting cleanup pipeline using geometric constraints and lineage pruning that matched segmentation-based isolation quality, then evaluated a ReAct agent for autonomous scene cleanup.
The project turns natural-language household scenarios into coordinated multi-application traffic for realistic, measurable CPUC network experiments.
I developed a CPUC traffic generator from natural-language queries and a cross-service observability system that traced execution, latency, and intent, reducing mean time to root cause by 8×.