MPhil researcher at the University of Sydney building reliable, efficient AI systems across federated learning, semantic communication, and trustworthy machine learning.
Distributionally robust semantic transmission for large-model inference across noisy, bandwidth-limited, and shifting wireless environments.
Contributed modelling, experiment design, robust training simulations, and evaluation under constrained channels.I work across the full loop: identifying a reliability gap, designing an algorithm, building controlled experiments, and turning the result into a reproducible system.
From controlled evidence feedback to distributional robustness and federated equilibrium learning.
Experiments are designed to explain when a system is right, how confident it should be, and where it breaks.
Reliable pipelines, constrained networks, maintainable interfaces, and artifacts built for repeatable use.
The immersive layer is optional. These case files keep the role, method, evidence, and links immediately scannable on every device.
Distributionally robust semantic transmission for large-model inference across noisy, bandwidth-limited, and shifting wireless environments.
Contributed modelling, experiment design, robust training simulations, and evaluation under constrained channels.Federated deep equilibrium models for heterogeneous clients where memory, energy, data, and communication capacity vary across the network.
Implemented FeDEQ components, explored communication-efficient aggregation, and evaluated non-IID NLP and vision settings.Reliable label aggregation under sparse crowdsourcing by separating block-local and global posterior structure and controlling recursive evidence feedback.
Zerun Niu — first author; led algorithm design, literature review, experimental design, code implementation, and experimental deployment.A maintainable public research interface for the Distributed compUting, optimizAtion, and Learning group at the University of Sydney.
Designed, developed, deployed, and maintain the group's official web presence and research communication system.Statuses are stated exactly: published, submitted, or under review. First-author BRAVE work remains explicitly marked as under review at TMLR.
Zerun Niu, et al.
L. T. Le, S. H. Wanasekara, Zerun Niu, et al.
L. T. Le, Zerun Niu, T. D. Nguyen, et al.
My research practice is strengthened by teaching: precise explanations, careful feedback, and systems that help people reason through complexity.
The University of Sydney
UNSW Sydney
The University of Sydney
DUAL Group, The University of Sydney
I’m Digital Zerun, an AI representation using Zerun’s authorised cloned voice. I answer only from a screened public knowledge file and cannot make commitments on Zerun’s behalf.
Voice activation awaits the owner’s Cloudflare and ElevenLabs account setup. Verified static answers and all site navigation remain available now.
No matching route. Try “BRAVE” or “CV”.