Haichang Li

HelloWorld ✍

I am Haichang(Charles) Li, a researcher and builder of. Previously, I was a founding member of Sounding.AI, which is a startup raising 10M+ CNY to explore AI Agent system, and I also worked with Flowtica and MiraclePlus (Formerly YC China), and collaborated with MEMO. During my undergraduate studies, I was advised by ACM Distinguished Scientist/IEEE Fellow Prof. Yung-hsiang Lu. My experiences span real-world and academia, which have shaped my passion for making something people want and a research focus oriented toward practical applications and productization ('From Paper to Product') in THE real world.

My work centers on Human-AI Collaborative Intelligence, exploring how people can construct the next generation of living with AI by transforming the world from being merely online to becoming OnAI. In this vision of Human-AI collaboration, individuals and all forms of online data become nodes within a new network grounded in AI. These nodes are connected not only through the internet but also through AI itself—whether in interacting with data or other users — positioning AI as the foundational infrastructure of this emerging paradigm.

My research explores how to design the OnAI symbiosis of Human-Agent Collaboration across three dimensions:

（a）Human - Understanding user needs through qualitative and quantitative insights, and designing interaction models that allow people to transmit intent and construct context, rather than being constrained by fixed pipelines.

（b）Agent – Building tools and designing intermediated artifacts that extend agent capabilities, enabling AI to align with human goals, persist across time and users, and act in both virtual and physical worlds.

（c）Collaboration – Rethinking the relationship between humans and agents so that AI becomes infrastructure for next-gen society, not just a tool — allowing people to truly live OnAI, much as we now live online.

My vision of "From Online to OnAI"

From Online to OnAI describes the next fundamental leap in our digital existence. Today, the world is predominantly online: people access data and interact with others through internet-mediated interfaces, relying on the existing network infrastructure. My research envisions a transformation beyond this paradigm, creating OnAI, where AI is no longer merely an application on the internet but an integrated layer within the global digital infrastructure.

In this OnAI paradigm, individuals, data, services, and other users all become nodes within a unified AI-grounded network. Connections between nodes are not limited to the Internet’s routing protocols—they are dynamically mediated by AI, enabling context-aware access, coordination, and collaboration. In other words, AI augments and fuses with existing digital infrastructure, becoming a foundational substrate that supports all flows of information, interaction, and activity.

Within this network, people can focus on being themselves, while AI orchestrates interactions with data, services, and others in real time, intelligently bridging gaps and simplifying connections. Just as the internet required entities to be “online” to participate in the digital ecosystem, everything in the future must be “OnAI” to fully engage with the AI-mediated network.

Works 📖

Research-based Exploration

• Memory as a Service (MaaS): Rethinking Contextual Memory as Service-Oriented Modules for Collaborative Agents Haichang Li Position Paper, arXiv
• A11yShape: AI-Assisted 3-D Modeling for Blind and Low-Vision Programmers Zhuohao Zhang, Haichang Li, Chun Meng Yu, Junan Xie, Gene S-H Kim, Faraz Faruqi, Mingming Fan, Angus Forbes, Jacob O. Wobbrock, Anhong Guo, Liang He. In Proceedings of the ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2025) Note: Leading researcher on system and study. Originally co-first author; authorship adjusted to support a final-year PhD’s graduation.
• Visualize Music Using Generative Arts Brian Ng, Samantha Sudhoff, Haichang Li, Joshua Kamphuis, Tim Nadolsky, Yingjie Chen, Kristen Yeon-Ji Yun, and Yung-Hsiang Lu In Proceedings of the IEEE Conference on Artificial Intelligence (CAI 2024) (Prior Survey for Mus2Vid)
• A11yReview: A Literature Review on Accessibility Artifacts Anonymous for potential review conflicts Manuscript in preparation for submission to CHI 2026.

Real-World Implementations

• Flowtica AI A voice-first agent and intelligent pen designed to transform AI from a single-turn tool into a long-term, collaborative partner. Visit Website ↗
• Shine Resume An early exploration into Multi-Agent LLM systems, built to help students with ordinary backgrounds stand out by providing AI-powered resume optimization and real-time interview aids. Visit Website ↗

Experience 💼