Fuyu Qiu

Experience

Multimodal Robot Dialogue Intern (Interaction Intelligence)

AgiBot (智元机器人)

Jun 2026 – Aug 2026

Built the head expression system for the AgiBot A3 robot, rendering parameterized and semantically driven facial expressions in real time from the text the robot is speaking, so that its affect stays synchronized with live speech.
Developed the human–robot interaction agent for A3, covering long-horizon task planning, vision-grounded tasks, and real-time barge-in — the agent can be interrupted mid-task and re-enter the suspended plan once the interruption is handled.

Co-founder & Lead Software Engineer

OpenEd

Remote / Champaign, IL | Jan 2025 – Present

Architected and deployed a specialized AIGC pipeline that transforms natural language into precise STEM educational videos by using LLMs to generate structured animation code rather than stochastic diffusion frames.
Ensured 100% visual consistency and mathematical accuracy for complex scientific content, including dynamic coordinate systems, geometric proofs, and LaTeX-rendered equations, by orchestrating a Python-based visualization engine.
Built a scalable SaaS backend on Alibaba Cloud using Docker, Redis for task queuing, and Postgres to manage the reliable execution of high-concurrency rendering jobs.
Developed a custom orchestration layer in TypeScript to manage multi-step LLM workflows (Gemini, Claude), bridging the gap between raw AI output and production-ready visual assets.

Explain Integral

Explain Stack

Hook's Law

Undergraduate Research Assistant

Rehg Lab (Prof. Jim Rehg), UIUC

Champaign, IL | Aug 2025 – Present

Developing Multimodal Video Large Language Models (VLM) for Social AI applications, specifically focused on processing complex temporal video data to interpret human social behaviors.
Architecting large-scale data pipelines and fine-tuning vision-language models to bridge the gap between raw visual features and high-level semantic behavior understanding.

Software Engineering Intern (Full-Stack & Vision)

Stopto-Optics

Zhenjiang, China | May 2025 – Aug 2025

Developed an industrial-grade analysis suite using React, Electron, and Babylon.js, creating a rich 3D spatial environment to visualize high-precision optical surface data beyond 2D limitations.
Bridged low-level hardware communication with the Web DOM to enable real-time, low-latency feedback loops between physical optical sensors and the user interface.
Optimized custom YOLOv11 object detection models using ONNX Runtime for high-efficiency inference on ARM-based embedded processors like Raspberry Pi.
Implemented a robust remote Over-The-Air (OTA) update system using C#, ensuring seamless deployment and reliable version control for software running on industrial edge devices.

Mobile Full-Stack Developer Intern

Olinteach

Shanghai, China | June 2024 – Aug 2024

Engineered a serverless Node.js backend to orchestrate AIGC pipelines, ensuring stable API performance and low-latency delivery for an automated video synthesis application.
Collaborated in an Agile environment to integrate Cloud Media APIs for real-time digital avatar replacement and video post-processing workflows.

Volunteer Developer

Consoul-in Mental Health Initiative

2024

Built an online therapist–patient communication platform supporting remote mental health counseling.

University of Illinois Urbana-Champaign

B.S. in Computer Engineering

Champaign, IL | Expected Graduation: May 2029

GPA: 4.0/4.0

Projects

inklet — Ambient E-Ink Displays for Your Second Brain

Connected Hardware & Cloud Platform | 2026 | Website | LinkedIn

inklet builds connected e-ink displays that bring the right information into the rooms where you live and work — contextual rather than chaotic, with no backlight and no notifications to fight for attention. The D1 display and the H1 Compute Hub sync notes, schedules, and documents from Notion, Craft, and Obsidian, and AI content routing decides which screen in which room a given piece of content belongs on.

On the software side, Portal is the cloud dashboard that manages every display from one place, and the Portal SDK (@inklethq/sdk) reduces publishing to a single typed call: hand push.auto() text, a link, an image, or a PDF, and inklet handles layout, room selection, and panel rendering. Content can run through Portal cloud or stay entirely on a self-hosted Compute Hub, keeping the system private by design.

Positron — Runtime JS Injection Toolkit for Electron

Systems & Reverse Engineering | 2026 | GitHub

Positron attaches to any running Electron application on Windows and evaluates JavaScript inside its main or renderer process — without restarting the target or modifying its files. A C++ host manually maps a payload DLL into the process, resolves V8 symbols through fuzzy PE export matching, and bootstraps a JavaScript server on the V8 main thread; the payload then self-unmaps, so nothing native is left in memory and later sessions reconnect straight to the live JS server instead of re-injecting.

The toolkit ships an interactive REPL that switches between the main and renderer worlds, a module system with lifecycle hooks for hot load and unload, one-shot command execution, x64 and x86 support, and a C++ SDK for programmatic integration.

A3A Protocol - Agentic AI & Web3 Commerce Platform

ETHGlobal ETHOnline 2025 | ETHGlobal Showcase | LinkedIn

Built the A3A Protocol enabling secure, real-time communication and payments between AI agents on Web3. Developed a Next.js-based backend for an autonomous commerce system that manages the entire order lifecycle, including logistics, payments, and status verification through a conversational interface. Integrated smart contracts and deployed on the ASI Agentverse for Fiducia Agents.

Implemented function calling to dynamically translate natural language prompts into executable database operations and blockchain transactions, enabling seamless agent-to-user interaction.

AI Play Edge Browser's Surf Game

Reinforcement Learning | 2025 | GitHub

Trained an AI model using reinforcement learning techniques to autonomously play the Edge browser's surf game. The model learns optimal strategies through trial and error, achieving high scores consistently.

LLM-Driven Interactive System for Accessibility

AI Research Project | 2024

Partnered with clinical psychology experts to design an inclusive educational tool for children with ASD, focusing on interactive scenarios to practice social navigation. Engineered an adaptive learning system featuring LLM-driven question generation, a responsive mobile frontend, and a backend analytics engine to monitor and visualize student progress metrics.

MBTI Personality Classifier Chatbot

Deep Learning & NLP Application | 2024 | GitHub

Developed an interactive application that performs multi-label text classification on WeChat chat logs using a fine-tuned BERT model to infer MBTI personality types. Designed a preprocessing pipeline to handle unstructured social media data, implementing custom tokenization and context extraction techniques to improve model inference accuracy.

A Simple Demoscene

Graphics Programming | 2025

Created a demoscene program in less than 4KB size, featuring animated procedural 3D objects and procedural music. Demonstrates advanced techniques in code optimization and real-time graphics rendering.

Casio Calculator Reverse Engineering

Security Research | 2023 | GitHub | Simulator

Discovered and exploited a stack overflow vulnerability in Casio fx-991CNX, enabling execution of arbitrary code using Return-Oriented Programming (ROP). Published findings on Zhihu and GitHub, fostering a research community around calculator security.

Documentation: Part 1 | Part 2 | Part 3

Custom x86-64 Operating System (QNOS)

Systems Programming | 2022 | GitHub

Developed an operating system from scratch in C, implementing Virtual Memory Management (VMM), Virtual File System (VFS), and multitasking capabilities. Demonstrates deep understanding of low-level systems programming and computer architecture.

Quick-C Compiler

Compiler Design | 2025 | GitHub

Building a C-like language compiler entirely in C, focusing on speed and minimal dependencies. The project emphasizes clean code generation and efficient parsing techniques.

Email:	qiufuyutony (at) gmail (dot) com
Find me on:	GitHub \| LinkedIn \| Website
Address:	409 E Chalmers St., Champaign, IL 61820
Resume:	Download PDF

2026.08	Completed the Interaction Intelligence internship at AgiBot (智元机器人), building the facial expression system and the human–robot interaction agent for the A3 robot.
2026.07	Announced inklet, connected e-ink ambient displays with a cloud Portal and a TypeScript SDK.
2026.05	Released Positron, a runtime JavaScript injection toolkit for Electron applications on Windows.
2025.01	Co-founded OpenEd, developing AIGC pipeline for STEM educational video generation.
2025.01	Started as Undergraduate Research Assistant at Rehg Lab, working on Multimodal Video Large Language Models for Social AI applications.
2024.08	Completed Software Engineering Internship at Stopto-Optics, developing industrial-grade analysis suite with React, Electron, and Babylon.js.
2024.06	Completed Mobile Full-Stack Developer Internship at Olinteach, engineering serverless Node.js backend for AIGC pipelines.