Photo of Jaber Daneshamooz

Jaber Daneshamooz

PhD student in Computer Science at UC Santa Barbara, advised by Arpit Gupta in the Systems and Networking Lab.

Research

I build systems that generate realistic, controllable data for networking machine learning, so that models trained in the lab still work in the field.

NetReplica (NTC-Reloaded 2026) recreates challenging network conditions by separating a bottleneck link's capacity and queuing from the competing traffic that crosses it; training a video-streaming (ABR) model on this data cut its prediction error by up to 47%. NetGent (NeurIPS ML4Sys 2025) uses AI agents to automate real application workflows across more than 50 applications. Pramana (HotNets 2026) ties them together: a researcher describes an experiment in natural language and Pramana runs it and produces the data.

I also work on foundation models for network traffic (netFound, NetBurst) and on teaching networking in the age of AI. I have mentored 13 students and founded NetSplit, a startup applying this work to internet management in shared environments.

News

Publications

Full list with all co-authors: Google Scholar and the CV.

Selected honors