Agentic Lab: An Agentic-physical AI system for cell and organoid experimentation and manufacturing

Area of Science:

• Artificial Intelligence in Life Sciences
• Bioengineering and Biomanufacturing
• Computational Biology and Bioinformatics

Background:

• Biological research and manufacturing face reproducibility challenges due to complex protocols and data analysis.
• Current experimental workflows are often static, limiting dynamic adaptation and learning.
• Integrating AI into the physical laboratory environment is an emerging frontier.

Purpose of the Study:

• To introduce Agentic Lab, an agentic-physical AI platform designed to enhance reproducibility in biological research.
• To bridge the gap between AI-driven reasoning and real-world laboratory operations.
• To create a collaborative and adaptive experimental lifecycle integrating AI agents and human scientists.

Main Methods:

• Utilizes a multi-agent orchestration architecture with specialized subagents for knowledge retrieval, protocol design, and multimodal data analysis.
• Employs large language models (LLMs) and vision language models (VLMs) for reasoning and phenotyping.
• Integrates an augmented reality (AR)-based physical AI interface for real-time monitoring and error correction during experiments.

Main Results:

• Agentic Lab autonomously generated protocols and monitored organoid differentiation from human pluripotent stem cells.
• The platform identified subtle morphological heterogeneity linked to growth conditions and interpreted phenotypes.
• Real-time error identification and context-aware instructions were provided to human operators, improving differentiation efficiency.

Conclusions:

• Agentic Lab transforms biological experimentation into an adaptive, feedback-driven process by unifying agentic AI with physical laboratory awareness.
• The platform facilitates dynamic collaboration between scientists and AI agents, closing the loop between planning, action, and analysis.
• Establishes a foundation for intelligent laboratories that integrate design, execution, and interpretation within a unified agentic-physical system.