About Me
Hi! I’m a machine learning engineer/researcher focused on building ML systems for real-world decision-making problems in product and enterprise settings. I am currently open to new opportunities to deepen my experience in designing ML-driven decision systems across domains such as fintech, insurance, supply chain optimization, and consumer products, where the key challenge lies in problem formulation and end-to-end system design rather than model optimization alone.
Previously, I worked at UC Capital as a machine learning researcher/engineer, where I built production-grade ML systems in a trading environment. I was involved in the full system stack, including data pipelines, multi-threaded infrastructure, interprocess communication, and latency-sensitive deployment. The role also required close collaboration with traders in a fast-moving and ambiguous environment, translating domain heuristics into machine learning systems under real-world constraints.
Prior to joining the industry, I led research in reinforcement learning and deep learning at National Taiwan University, resulting in publications at ICML and NeurIPS. My work was supported by the NSTC international travel grant., which enabled me to present my research in Vienna and Vancouver. My research spans reinforcement learning, imitation learning, diffusion models, and decision-making systems.
Outside of work, I enjoy building LEGO sets, snowboarding in winter, and recently started learning tennis, still early but enjoying the process.
Interests
- Deep Learning
- Computer Vision
- Generative Models
- Robotics
- Reinforcement Learning
- LLM
Education
-
M.S. in Communications Engineering, 2024
National Taiwan University
-
B.Eng in Information Engineering, 2022
Shanghai Jiao Tong University
Publications
Experience
Machine Learning Engineer
Independent
Key Contributions:
- Built a scenario-based framework that translated geopolitical events into portfolio risk signals through causal analysis.
- Developed a reinforcement learning system for laptop thermal control, improving system efficiency under dynamic workloads.
Machine Learning Quantitative Researcher
UC Capital
Key Contributions:
- Spearheaded the redesign of an institutional trade execution system, ultimately developing a risk-controlled statistical decision framework that improved average execution price by 1 basis point (bp), ≈ USD 1.3M annual profit uplift.
- Established the company’s first trusted market data foundation by validating and standardizing over 10.8TB of tick-level trading data (2 years, 1,800 stocks), replacing 12+ fragmented datasets with a centralized source adopted across trading and AI teams.
- Developed an interpretable limit-up stock ranking system by transforming 5 trader heuristics and adding 8 features, reducing stock selection time from 2 hours to 30 mins, improving next-day returns by 0.5%, yielding USD 30M+ annual trading impact.
Master Research
RLLab in NTU
Key Contributions:
- Joined Prof. Shao-Hua Sun's research group, focusing on deep learning and reinforcement learning for sequential decision-making in robotics.
- Conducted research in imitation learning and generative decision-making, focusing on how diffusion models improve behavioral cloning and adversarial imitation learning under high-uncertainty environments.
- Led the development of diffusion-based imitation learning frameworks (Diffusion Model-Augmented Behavioral Cloning and Diffusion-Reward Adversarial Imitation Learning), covering full research cycles from hypothesis formation, modeling design, and empirical validation to publication at ICML 2024 and NeurIPS 2024.
- Identified and resolved fundamental modeling limitations through first-principles analysis, leading to a successful pivot of a stalled research direction into a NeurIPS 2024 publication in collaboration with a senior NVIDIA researcher.
- Expanded the lab’s research direction toward diffusion-based generative modeling for reinforcement learning and imitation learning, inspiring multiple follow-up projects across academia and industry collaborations.
Undergraduate Thesis
Department of EE at SJTU
Key Contributions:
- Research ideas for identifying actions in tennis matches.
- Develop software to extract essential moments from game footage.
Undergraduate Research Intern
Lab of Prof. Jiaxin Ding
Key Contributions:
- Explored solutions to find embedding vectors in GPS trajectory data, and validated the approach through implementation.
Engineering
SJTU Autonomous Driving Team
Key Contributions:
- Led development of a LiDAR-SLAM and lightweight YOLO-based perception system for an autonomous racing vehicle, enabling real-time obstacle avoidance under strict embedded constraints and improving detection performance from 60% to 88%.
- Designed a closed-loop data iteration pipeline (logging, relabeling, retraining) and coordinated system-level integration across sensors, hardware constraints, and latency budgets, ensuring stable end-to-end autonomy performance.
- Selected as team lead for technical excellence; supervised junior members and conducted iterative testing to optimize performance.
Tools include:
- C++, Python, YOLO, SLAM, ROS
Engineering
SJTU Robomaster Team (云汉交龙)
Key Contributions:
- Contributed to real-time perception and control systems for robotic armor detection and targeting under strict competition constraints.
- Diagnosed instability in auto-aim behavior by decoupling perception from sensor calibration issues, revealing root causes in the vision pipeline.
Tools include:
- Python, Computer Vision, OpenCV
Contact
- hc.nathan.wang@gmail.com
- Follow Me
- Connect On