Research collaboration with Northwestern

A collaboration in embodied AI

I began a remote research collaboration with Prof. Manling Li’s group at Northwestern University, mentored by Qineng Wang. The project explores implicit task affordance prediction in 3D space.

Connecting language to possible actions

The research links videos, point clouds and camera trajectories to implicit natural-language task instructions, with a focus on navigation-related affordances.

My work includes language conditioning for PointFormer-style backbones and an early data-curation pipeline in which vision-language models annotate videos with per-frame subtasks and affordances.

From prototypes to navigation

I also work on the transition from VLM-only prototypes toward a navigation model conditioned on language and affordance signals. This collaboration extends my agent research toward grounded perception and action in spatial environments.