Previously, I completed robotics internships and co-founded a robotics startup, focusing primarily on SLAM.
I build persistent spatial representations for embodied agents, drawing on SLAM, lifelong mapping, 3D scene understanding, and egocentric perception.
My current research explores how wearable and robotic agents can maintain geometrically grounded, semantically structured, and continuously updateable memories of dynamic environments.
Research
Selected Research
LEXI-SG: Monocular 3D Scene Graph Mapping with Room-Guided Feed-Forward Reconstruction
Builds scalable, open-vocabulary 3D scene graphs from monocular RGB by combining room-level hierarchy, feed-forward reconstruction, and factor-graph alignment.
Christina Kassab*, Hyeonjae Gil*, Matías Mattamala, Ayoung Kim, and Maurice Fallon
IEEE Robotics and Automation Letters (RA-L) · 2026
Fieldscale: Locality-Aware Field-based Adaptive Rescaling for Thermal Infrared Image
Uses locality-aware two-dimensional fields to rescale raw thermal images with spatially adaptive contrast, improving visibility and downstream perception.
Hyeonjae Gil*, Myung-Hwan Jeon, and Ayoung Kim
IEEE Robotics and Automation Letters (RA-L) · 2024