Hao Zhang

Worked on deploying Large Language Models (LLM) on power-constrained devices. Focused on quantization and benchmarking platforms for evaluating LLM performance on mobile devices.

Utilized deep learning algorithms to simulate defense and attack processes for IoT devices and detect data exfiltration. Developed and deployed a flow size predictor using TCP packet datasets to optimize networking system performance.

Developed and deployed AR applications on wearable devices, including face detection on VR headsets with eye-tracking integration. Explored and replicated state-of-the-art real-time speech separation baselines, including simulating synthetic audio datasets for Beamforming validation.

Participated in the NeurIPS 2019 MineRL Competition, achieving 4th place. Focused on distributed reinforcement learning, combining PPO with Vtrace to enhance performance in MineCraft and Impala reproduction.

Contributed to building a deep learning system for graph-based models and worked on graph-based model benchmarks. Proposed algorithms for sparsifying molecule datasets based on edge attention.

Activities and societies: Engaged in research on deploying LLM on power-constrained devices, including quantization and benchmarking platforms for evaluating LLM on mobile devices. Contributed to various research papers presented at conferences like ICLR and AAAI.

Currently pursuing a PhD in Computer Science, supervised by Prof. Suman Banerjee. Research interests include Large Language Models (LLM), Graph Representation Learning, and Wireless Sensing.

Completed a Bachelor's degree in Electronic Engineering. Gained foundational knowledge in electronic systems and related technologies.