About · Zhiway
Amplify scientists,
don't replace them
Zhiway is an AI for Science company, on a mission to build the scientific infrastructure for the agent era. We believe research output isn't limited by a researcher's intellect, but by how much repetitive work dilutes it — so we hand nine-tenths of that work to agents, freeing people to focus on the judgment and creativity nothing else can deliver.
Now that the ability to finish a given task can be bought by the month like any commodity, the mainstream of AI for Science has set out to replace the scientist. Zhiway chose the opposite path.
Why we exist
What research truly lacks has never been the ability to "write" — it's trustworthy delivery across hundreds of steps: can the AI's output be attributed by name, can every source be verified one by one, can long-horizon tasks be safely entrusted to it. None of this emerges on its own from general-purpose capability; it can only be purpose-built for serious research.
So we use agents as a glue layer, weaving together the literature, code, experiments, writing and communication a researcher touches every day; we ground verifiable citations on a 150TB foundation of full-text papers; we take on long-horizon tasks through multi-agent orchestration; and we insist on a triple gate of real data, traceable process and verifiable output.
How we named ourselves
The Chinese name 清智万维 carries three characters: Qing, from Tsinghua University; Zhi, for artificial intelligence; Wanwei, the world-wide web of ideas, citation networks and creative connections. Together they capture exactly what we care about most: using an academic pedigree and the power of agents to accelerate how ideas flow and collide across the citation network.
Where we're headed
First, organize literature, code, experiment data and writing into a trustworthy scientific operating layer. Next, enter the physical layer asset-light: connect instruments with open APIs first, then use controlled UI access and robot adapters for broader compatibility. VLA and robot vendors own action generation, the body and low-level control; Zhiway connects protocols, experiment rules, human approval, success criteria, failure replanning, hard stops and evidence traces. This direction is in technical reserve and pre-research, with validation planned for late August 2026.
Journey
From a service into a research operating system
It began in 2024 as a research-outsourcing service — proof that people would pay, yet capped at "delivery equals two founders' time." We codified that judgment into a system, and a single service grew, over time, into four products on one shared engine — now reaching toward the physical layer.
Team
Covering precisely every capability this work demands
First-hand research experience, multi-agent engineering, and a channel into institutional and international markets — a team that holds all three at once is, in itself, a scarce asset.
Siyuan Liu
Founder · CEO
- MSc in CS / AI, University of Southampton; large-model algorithm expert
- 10+ years bringing AI applications to production; five products built solo
- Led the development of China's first medical LLM and a financial-research LLM
Xueqi Zhao
CSO · Chief Strategy Officer
- Digital and AI policy expert at the United Nations Industrial Development Organization (UNIDO)
- PhD; Assistant Researcher, School of Economics and Management, Tsinghua University
- Has led national and provincial/ministerial projects and published 15 high-level papers; gold medal at the National Innovation & Entrepreneurship Competition
Yilong Li
Co-Founder · Head of Engineering
- Tech lead for the Xiangyu product line at 5i5j Group
- SaaS expert and full-stack Java web engineer
- Has co-founded LLM and agent startups with Siyuan Liu since 2023
Fei Meng
Co-Founder · Head of Operations
- Honours BEng in Engineering, University of Adelaide, Australia
- Project manager at Stantec; five years of full-lifecycle project management in international consulting
- Delivered projects with a combined budget of over AUD 10 million; leads strategic partnerships across Asia-Pacific
Haonan Zhang
Head of University–Industry–Research Partnerships
- PhD and Shuimu Scholar; Assistant Researcher, Institute for AI International Governance, Tsinghua University
- Member of the Academy of Management (AOM)
- 7 papers in leading international journals and 4 at top conferences; national gold medal at the Challenge Cup
Jinghui Yin
Head of Investment & Financing Partnerships
- PhD; Assistant Researcher, School of Economics and Management, Tsinghua University
- 10 high-level papers published
- Five research reports received high-level endorsement or adoption
With the help of Zhiway's system, researchers have authored and put their own names to papers that have earned 3 acceptances at IJCAI, EMNLP and AAAI, with a fourth under review — and blind review at top conferences is the harshest third-party test this field has.
Building the scientific infrastructure for the agent era together
Serving the next generation of scientific discovery and experimental R&D through research-workflow agents, experiment-grade feedback loops and asset networks.
Building the Scientific Operating Layer for AI-Driven Research
Research-Workflow Agents
Experiment-Grade Feedback Loop
Scientific Infrastructure Platform
FAQ
About Zhiway
What kind of company is Zhiway?
Zhiway (legally Zhiway (Xiong'an) Technology Co., Ltd.) is an AI for Science company building the scientific infrastructure for the agent era. Our philosophy is simple: amplify scientists, don't replace them — we hand roughly nine-tenths of a researcher's repetitive work to agents, so people can focus on the judgment and creativity that nothing can replace.
What does the name "Zhiway" mean?
In Chinese, the name 清智万维 reads as: Qing, from Tsinghua University; Zhi, for artificial intelligence; Wanwei, the world-wide web of ideas, citation networks and creative connections. The English name Zhiway echoes "the way of intelligence."
What products does Zhiway build?
The product matrix shares one scientific operating layer: Idea Network is agent-native research-graph infrastructure; Idea Interflowing is the graph-grounded workspace; ResearchClaw is a local research operating system; ThesisAgent supports standardized research writing; and HashMind preserves reusable agent assets.
How is Zhiway different from other AI for Science companies?
Most AI for Science roadmaps set out to replace the scientist; Zhiway goes the opposite way — we amplify them. We commit to trustworthy delivery built on real data, traceable process and verifiable output, and we serve institutional clients with local-compliance capabilities (data never leaves the premises, with domestic-stack adaptation).
How can I contact Zhiway?
Email us at contact@zhiway.com.cn.
Walk with us
Whether you're a researcher, an institution, or simply someone who cares about AI for Science — we'd love to hear from you.