A human direction for AI.

AI Certainty is a founder-led research and product initiative. Our long-term aim is to make human state a first-class input to intelligent systems, while turning research into reusable foundations.

Starting point / Long-term outlook

From Shanghai,
outward.

Shanghai is our starting point. Our long-term ambition is to build research and tools for an international community.

Shanghai, ChinaInternational outlook

Proposed company name / 拟登记名称

上海确知智能科技有限公司

The proposed name is subject to approval. The company has not yet been registered.

Junjie Xu

Junjie Xu is a Direct Ph.D. researcher at East China Normal University, School of Computer Science and Technology, Shanghai, working on multimodal affective computing and generative systems.

Research statement
“I want to place cross-modal alignment, affective understanding, and controllable generation on the same research path.”

His publication record spans multimodal generation, affective AI, and language understanding: 14 first-author papers among 27 publications. Two ECNU research projects led, one rated Excellent at completion. 14 published Chinese patent applications as co-inventor — publication numbers shown for verification.

His recent work turns to evidence and evaluation fidelity: auditing whether LLM safety judgments follow expressed evidence or severity cues (Evidence Before Severity, Under review · AAAI 2027), and separating changes in agent behavior from changes in evaluation rules (ANI — Submitted · ICLR).

From November 2026 to November 2027, Junjie Xu will be a CSC-funded visiting researcher at Nanyang Technological University, hosted by Prof. Eric Cambria. The placement is personal academic funding for the founder; it is not an affiliation, endorsement, or partnership of AI Certainty.

Student awards include the 2025 Kuanrui Talent Scholarship and the 2023 Shanghai Outstanding Graduate award.

Personal website Google Scholar GitHub ORCID

Not yet

  • Not a registered company — no legal entity is claimed on this site.

  • Not a team — the initiative is currently one person.

  • Not a product — no users, deployments, or uptime exist to report.

  • Not funded as a company — there are no investors to announce.

When any of that changes, this page changes with it.