OPEN RESEARCH · COMPOSABLE CAPABILITIES · EVIDENCE FIRST · RESEARCHER LED
πAI Lab is a public research and open-technology initiative based at 广东智慧医学国际研究院 in Guangzhou. We build distinct but composable scientific capabilities across the research process—not a closed, all-in-one agent.
Development status: All directions below are under development. This page describes research scope and intended relationships; it does not imply public release, production readiness, adoption, or scientific, clinical, legal, or regulatory validation.
| 01 · Data & methods | OmniData · OmniEngine Agent-native scientific data and reusable methods with provenance, applicability conditions, validation, resource boundaries, and explicit failure returns. |
| 02 · Evidence & intelligence | OmniScholar · OmniPatent · AI4SNews Literature, patents, full text, figures, citations, research signals, retrieval, and claim-to-evidence relationships. |
| 03 · Continuous research | OmniMind · OmniHarness Research state, agent coordination, tools, permissions, observability, verification, recovery, and handoff across long-running scientific work. |
| 04 · Scientific artifacts | OmniPlotter · OmniSketch · OmniSlide · OmniOffice Scientific plots, editable illustrations, presentations, documents, spreadsheets, and reliable file operations through OmniDoc, OmniSheet, and PPT Skill. |
| 05 · Question formation | OmniSage Testable scientific questions formed from evidence gaps, competing explanations, and explicit validation paths. |
| 06 · Domain research | Drug discovery A demanding biomedical setting connecting domain data, scientific methods, agent workflows, evidence tracing, and reporting. |
These directions have different scientific jobs and are intended to remain independently reusable. Their shared goal is composability: researchers and agents should be able to move across evidence, data, methods, execution, and scientific artifacts without silently losing provenance, conditions, state, or failure information.
OmniMind is one environment in which these capabilities are intended to converge around a continuing research question. It does not own them or make itself their exclusive gateway; the underlying capabilities are intended to remain reusable by other agents and research platforms through open, explicit interfaces.
OPEN — Capabilities are not locked to one agent. They may be coordinated through OmniMind or reused by other agents and research platforms.
VERIFIABLE — Evidence travels with the work. Sources, versions, permissions, environments, limitations, and failures should remain inspectable.
RESEARCHER LED — Scientists retain the final judgment. Direction, interpretation, validation, and release remain human responsibilities.
Evaluation and reproducibility apply across the system: benchmarks, provenance, failure analysis, recovery, reproducible environments, and validation in real research settings.
Zaoqu Liu (刘灶渠) — Lead
We welcome research collaborations around scientific data and methods, evidence systems, agent architecture, scientific communication, evaluation and reproducibility, biomedical research, and cross-disciplinary transfer.
Contact: liuzaoqu@163.com
Public projects enter πAI Lab only when they have a clear scientific purpose, accountable maintainers, reproducible entry points, explicit licensing, documented evidence boundaries, and a credible maintenance path.
Governance · Project policy · Contributing · Security · Support
中文摘要
πAI Lab 是设在广东智慧医学国际研究院的公共研究与开放技术计划。我们从真实生物医学研究出发，计划建设可以跨模型、跨 Agent、跨课题持续复用的开放科学能力，而不是一个封闭的端到端科研智能体。
- 数据与方法： OmniData、OmniEngine
- 证据与科研动态： OmniScholar、OmniPatent、AI4SNews
- 持续科研与可靠执行： OmniMind、OmniHarness
- 可编辑科研产物： OmniPlotter、OmniSketch、OmniSlide、OmniOffice
- 科学问题形成： OmniSage
- 领域研究： 药物发现
这些方向各自承担独立的科学任务，目标是通过开放接口彼此组合。OmniMind 是它们可以汇聚的一种持续科研工作环境，但不是其他能力的所有者或唯一入口。评测、溯源、失败分析与可复现性贯穿所有方向；科学家始终负责研究方向、证据判断、实验验证和最终结论。
以上方向均处于建设中。列入本页不代表已经公开发布、达到生产状态、获得规模采用，或通过科学、临床、法律与监管验证。
科研合作：liuzaoqu@163.com
