Hejun Huang  

I am currently a Software Engineer at Amazon, where I'm responsible for a large-scale online recommendation and distribution platform in the Prime Video Core Tech team. Before joining Amazon, I worked as a research engineer at Baidu USA in collaboration with Mr. Liu He and Dr. Xin Zhou, focusing on innovative text-to-video projects at Baidu Research Institute. Earlier in 2024, I served as a Research Engineer at the Robotics and Autonomous Driving Lab under the mentorship of Dr. Liangjun Zhang. Prior to that, I earned a fully funded Master's degree in Aerospace Engineering from the University of Michigan, Ann Arbor, working with Prof. Max Z. Li.

My research spans large language model, machine learning, reinforcement learning, control theory and optimization, focusing on the development of Gen AI with reliable learning algorithms for cognitive intelligence.

Before my time in Michigan, I received my M.S. from The Chinese University of Hong Kong (CUHK). I had the privilege of being advised by Dr. Dongkun Han in the Department of Mechanical and Automation Engineering, where I also collaborated with Mr. Zhenglong Li and Prof. Vincent W. L. Tam. My undergraduate studies were completed in Mechatronics Engineering at North China Electric Power University.

From 2020 to 2021, I served as a full-time research assistant at CUHK, and was promoted to a research associate the following year, holding the position through 2022. In 2019, I interned in the Intellectual Property Department at Daimler Greater China Investment Co., Inc.

CV  /  Google Scholar  /  LinkedIn  /  GitHub

We built the first multimodal agent-based video generation pipeline through 3D engine scripting!

I developed Auto Cement Truck on the Apollo platform via C++ and Python, customized planning and control, calibration and perception modules for both open and closed-loop deployments using 92GB data within 4 months.

We propose the metrics termed Plausibly Reachable Zone (PRZ) and Envelope (PRE), which are designed to estimate the privacy score of a local region using dynamic data.

We propose RIPO, a framework integrating both RL-based trading agents and BF-based risk controllers to achieve high long-term profits under acceptable short-term risks.

We propose an idealized analysis of the coverage proportion within the Remote Identification systems for uncrewed aircraft systems, numerical simulations based on San Diego's urban dataset are executed to verify the result.

We combine Sum-of-Squares programming with Reinforcement Learning to generate real-time controller for diverse robotic applications.

We use a Region-of-Attraction guided controller, to stabilize the partially unknown dynamical system approximated by Gaussian Processes. Strict probability statement is assessed to guarantee stabilization.

We introduce a Lyapunov-Barrier function into a partially unknown system and estimate its certified Region-of-Attraction, to assess the effect of noises on stability.

• Summer Research Project, UMich, 2023.
• Conference Reviewer: CCTA'22.
• Conference Volunteer: ACC'22.
• Teaching Assistant: ENGG1910 Demystifying AI, CUHK, 2022.
• Summer Research Project, CUHK, 2021, 2022.

• Arts: I've been passionate about band shows and practicing the saxophone for over a decade. During my undergrad, I played saxophone in Ni-Nan-Yi, a multi-ethnic and diverse band. Our band was honored with the 'First Prize in the Baoding City Band Competition.' Besides regular campus performances, we organized a charity concert and donated 7,000 CNY to students in Daliangshan, Sichuan.
• Sports: I like soccer, badminton, swimming and hiking.

Last Update: June 2024

Website Template