Sean Ding
Projects

Selected work, separated from the home page.

This subpage keeps the project details in one place so the home page stays lighter. The Homework Planner screenshots are in a carousel here, and the images are sized to fit naturally instead of getting cropped.

Layout change

The carousel uses horizontal scrolling with scroll snap and natural image height, so the screenshots keep their full frame instead of being forced into a shorter crop.

2 Player Battle Simulator

Game prototype

What it does

A game prototype focused on simple controls and quick play. It is a good example of the creative side of my work and how I approach interactive design.

A/D to move W to jump

Recovery-Augmented Imitation Learning on SO-101 Arms

Robotics research project

What it is

A summer research project testing whether adding recovery demonstrations, (human corrections recorded after intentional or accidental failures), to standard imitation learning data can make robot policies more robust. I'm training Action Chunking Transformer (ACT) policies with Hugging Face LeRobot on two Hiwonder SO-101 arms set up for leader-follower teleoperation (chosen for their low cost and intuitive documentation), comparing a baseline trained on clean demonstrations only, against a mixed dataset of roughly 75% clean and 25% failure-and-recovery sequences. My goal is to publish a paper to the Journal of Emerging Investigators.

Tasks being trained:

Colored object sorting Block stacking Switch activation

Key facts

Timeline: July – mid-August 2026 (6–7 weeks)
Platform: Hugging Face LeRobot + 2× Hiwonder SO-101 arms
Target: JEI submission
Evaluation: 20+ autonomous trials per policy, logged by failure type

Future projects

Where new work will go

Add another project here by copying one of the cards above, changing the title, description, and media, and linking to the live site if you have one.