I’m Owen Hu, a master’s student in Artificial Intelligence at Columbia University. My research and projects connect visual perception with sequential decision-making: 3D seismic segmentation, skeleton–language alignment, reinforcement learning, and decision-support systems.
I’m interested in how agents learn from interaction and build models of their environment. I’m exploring reinforcement learning, world models, and physical / embodied AI as directions for future research.
A selection of my work in perception, decision-making, and systems.
Browse more projects ↓
Data–knowledge–physics integration for subsurface characterization
Nov 2023 – Dec 2025 · project; paper under review
Led a four-student project connecting geological data governance, 3D numerical modeling, cave segmentation, and seismic inversion through a data–knowledge–physics framework.
Deep reinforcement learning in SUMO
Jun – Nov 2024
Framing traffic signal timing as a Markov decision process and studying a DQN-based controller in simulation.
A platform for resource allocation
Apr 2024 – Apr 2025
Connecting demand, inventory, and allocation workflows through a web platform for disaster relief supplies.
Coursework, independent studies, engineering tools, and competition projects.
Apr 2026 · documented experiments
Developing skeleton–text alignment on the Neuron baseline: six experiments spanning semantic prompt design, CLIP feature diagnostics, and dual-stream online prompt learning.
Spring 2025 · course project
A UAV detection and tracking pipeline for video files and live streams, with a Streamlit interface and visual outputs.
Exploring DQN for the traveling salesman problem through a custom environment, training code, and baseline-solving scripts.
Modeling League of Legends equipment choices as a budget-constrained decision process, with DQN and sensitivity analysis.
A small local utility for observing a selected conversation, preparing preset replies, and confirming sends.
Jul – Aug 2024
A C# symbol-table conversion tool connecting QGIS and MapGIS, developed during my Zhongdi Digital Technology internship.
A survey draft comparing reinforcement-learning formulations in landmark detection, object detection, tracking, and segmentation.
Hands-on segmentation work covering landslide recognition and VOC2012 experiments with FCN and U-Net models.
2023 · MathorCup
Freight-volume forecasting and network rerouting under depot closures; awarded First Prize in the 2023 MathorCup undergraduate division.
2023 · CUMCM
Modeling optical efficiency and optimizing heliostat layouts; awarded National Second Prize in the 2023 undergraduate modeling contest.
2024 · Challenge Cup
A team entrepreneurship project on freight-inspection applications, awarded Hubei Provincial Gold in the 2024 Challenge Cup.
M.S. in Artificial Intelligence · In progress
B.Eng. in Computer Science and Technology
Backend Development Intern
Developed C# tooling to translate QGIS marker, line, and fill symbols into MapGIS representations. Refactored symbol-class relationships and organized reusable transformations for visual and geometric properties.
Head of the Center
Coordinated academic activities and research/competition support, including the Future Technology Frontier Forum, peer exchange, and the college debate team. Organized planning and communication, and prepared competition guides and event articles connecting students with academic opportunities.
13th Challenge Cup · College Student Business Plan Competition
Team project: Qingche Shulu — digital radiographic inspection for freight logistics.
13th MathorCup Mathematical Modeling Challenge
Team project on freight-volume forecasting and logistics network optimization.
China Undergraduate Mathematical Contest in Modeling
Team project on heliostat-field layout optimization.
For conversations about research, projects, or potential collaboration, you can reach me by email.
