Shiqin Tong

I'm spending a gap year at the Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, advised by Prof. Mengdi Xu, where I work on memory mechanisms for embodied agents, especially spatial memory and its update mechanisms in robots. More broadly, I'm curious about how such memory modules relate to in-context learning in foundation models.

Before that, I earned an M.S. in Health Informatics from Cornell University, advised by Prof. Tapomayukh Bhattacharjee and Prof. Jose Florez-Arango; an M.S. in Artificial Intelligence from the National University of Singapore; and a B.S. in Software Engineering from Dalian University of Technology, advised by Prof. Xiangjie Kong and Prof. Feng Xia.

I like observing real user needs in tightly regulated settings and building improvements for them, be it products, algorithms, or systems. That drew me to Weill Cornell Medicine and to West China Hospital, where I was advised by Prof. Chongyang Wang.

Some things I've come to believe: Lessons So Far.

Email  /  LinkedIn  /  Github

Shiqin Tong

Research

My work spans memory for long-horizon embodied agents, assistive robotics, and AI for healthcare (Neurotech). Papers under review are shaded blue; those marked * are Neurotech.

Dex-X: Learning Visual-Tactile Dexterous Manipulation From Human Demonstration with Simulated Interaction
Under review, May 2026

Learning dexterous manipulation that fuses vision and touch directly from human demonstrations.

* (Neurotech) ReTell: LLM-Powered Task-based Learning for Neuro-Inclusive Education of Children with ADHD
Under review, January 2026

A task-based learning system designed for neuro-inclusive education of children with ADHD.

* (Neurotech) Social Determinants of Health and Telehealth Utilization for ADHD: A Cross-Sectional Study
Under review, January 2026  ·  AMIA 2025 Workshop

A cross-sectional study of how social determinants of health shape telehealth utilization for ADHD care.

Knowing When Not to Help Knowing When Not to Help: Active Estimation of Human Reachability for Just-Right Robot Assistance
Ziang Liu, Yunting Yan, Christy Sum Yu Cheung, Tailai Ying, Bodong Liu, Shiqin Tong, Alexander Orkwis, Katherine Dimitropoulou, Tapomayukh Bhattacharjee
Robotics: Science and Systems (RSS), 2026

A robot actively estimates what a person can and cannot reach, so it steps in only when help is actually needed, and holds back when it isn't.

FEAST mealtime-assistance system FEAST: A Flexible Mealtime-Assistance System Towards In-the-Wild Personalization  (Best Paper Award)
Rajat Kumar Jenamani, Tom Silver, Ben Dodson, Shiqin Tong, Anthony Song, Yuting Yang, Ziang Liu, Benjamin Howe, Aimee Whitneck, Tapomayukh Bhattacharjee
Robotics: Science and Systems (RSS), 2025
project page / arXiv / proceedings

A modular mealtime-assistance robot that care recipients personalize in-the-wild, using LLM-adapted behavior trees for safe, transparent, and adaptable feeding, drinking, and mouth-wiping.

Mobile Edge Cooperation framework Mobile Edge Cooperation Optimization for Wearable Internet of Things: A Network Representation-Based Framework
Xiangjie Kong, Shiqin Tong, Hui Gao, Guojiang Shen, Kaoru Wang, Mario Collotta, Ilsun You, Sajal K. Das
IEEE Transactions on Industrial Informatics (TII), 2020

A network-representation framework for optimizing mobile-edge cooperation in wearable IoT. (67 citations)

Industry

My industry work spans CT imaging engines, LLM API and MCP orchestration, and Kubernetes clusters. When teammates were swamped, I also helped out with front-end, back-end, mobile 😂

United Imaging Healthcare Software Algorithm Engineer · Full-Stack  ·  Mentor: Zheng Zhou
United Imaging Healthcare, 2024–2025

CT imaging engine and distributed LLM inference for high-concurrency serving.

NewYork-Presbyterian Hospital Software Algorithm Engineer (Intern) · Full-Stack
NewYork-Presbyterian Hospital, 2023–2024

Medical dialogue agents and real-time event-stream scheduling.

A*STAR Algorithm Engineer (Intern)  ·  Mentor: Lux Anantharaman
A*STAR, Singapore, 2021–2023

Ray-based parallel inference pipeline and diffusion-based anomaly detection.

NPDP NPDP (New Product Development Professional)
PDMA, 2025 · product & management
TOGAF TOGAF 9 Foundation / Certified
The Open Group, 2022 (Singapore) · enterprise architecture & agile
Tableau Tableau Desktop Specialist
Tableau, 2022
JLPT N2 JLPT N2
Japanese Language Proficiency, 2020 (Dalian)

Lessons So Far

What I've picked up outside of papers: how I think and work, the books I read, the tools I've built, and what I do away from the desk.

Not conclusions, just what I've learned so far.

Math and Fundamentals
Why I stopped fearing math, and why I went back to the basics.
Brain Load and Recovery
A late-night headache taught me to stop when my brain is tired.
Redesigning My System
Treating daily work like a training run, with Obsidian as the log.
Visualization
A 3D playground where I watch algorithms move before I read the math.
Reading
118 books · 4,956 notes · a small dashboard and the lines I marked

Most of all, I'm grateful to every PhD student who has mentored me: more than the papers, it's these everyday habits that they really taught me.

My Obsidian knowledge graph Obsidian knowledge graph (collection)
built with Obsidian

A vault of atomic Zettelkasten notes linked into a graph I actually think in: every idea is one card, wired to its neighbors.

n8n paper graph Self-growing note tree (integrate and discover)
built with n8n

An n8n pipeline that grows the vault on its own and scores how important each note is. A write-up about it drew 400+ comments.

Blender scene imported into a simulator Blender2sim Loader
built with Blender

A loader that pulls scenes from Blender into simulators and composes them on the fly, one-click generating whatever scene I need, with agents assembling the pieces.

Caregivers dataset poster Statistics
built with SPSS · poster (PDF)

Working with pure mathematicians, I found that when an LLM has no obvious signal to follow, the people with real insight can quickly point me to the one we should supervise.

Figure Skating

Favorite skater: Yuzuru Hanyu. I can skate a basic short program, and jumps are my favorite part. I learned them the plain way, falling and getting up until the take-off turned into a reflex with no hesitation.

Piano

I love pop songs arranged from classical music: pop written with classical methods. I also arrange music myself.


Template based on Jon Barron's website.