Jack Goler

Jack Goler

I'm an undergraduate researcher at Stanford University's Robotic and Embodied AI Lab (REALab), advised by Shuran Song. I work on robot learning for manipulation — most recently on underwater manipulation that avoids underwater teleoperation.

At Stanford I study mathematics and EE/CS. Before REALab I was at Stanford's Navigation and Autonomous Vehicles Lab, where I worked on photorealistic digital twins of environments using NeRFs and Gaussian splats trained on drone imagery.

I'm applying to PhD programs this fall.

Research

I'm interested in how robots acquire manipulation skills from data that is cheap to collect — human demonstrations, handheld grippers, and self-supervised interaction — and in how those skills transfer across embodiments and environments. I also have interests in graphics and systems engineering.

Publications

UMI-Underwater: Learning Underwater Manipulation without Underwater Teleoperation

Hao Li*, Long Yin Chung*, Jack Goler, Ryan Zhang, Xiaochi Xie, Huy Ha, Shuran Song, Mark Cutkosky

RSS 2026 *equal contribution

Project page Paper arXiv Video Code

Underwater manipulation is bottlenecked by the cost of teleoperated data collection and by how badly RGB policies degrade in water. We pair an autonomous, self-supervised underwater data collector with a depth-based affordance representation trained on on-land handheld demonstrations, which transfers underwater zero-shot. Deployed in the pool and in the ocean at Stanford's Hopkins Marine Station.

Explainers

Short videos where I rebuild an idea from scratch — the surest way I know to find the gaps in what I think I understand.

Bigram Language Modeling

Course Recaps

What I worked on each quarter at Stanford.

Spring 2025 Recap