The Aqua Project
Building a generalist control agent for every experiment on the planet
The ability to automate complex experimental processes is pivotal to scaling advanced technologies. Modern scientific experiments require precise control of complex systems, from optical cavities to quantum devices. Traditional control methods rely heavily on manual tuning and domain-specific expertise, creating bottlenecks in research and limiting experimental scalability. AQUA: A Quantized Utility Agent, represents a paradigm shift toward autonomous experimental control.
The research vision centres on developing AI agents that can learn to control any experimental setup with minimal human intervention. By combining generative world models with sample-efficient reinforcement learning, AQUA adapts to new environments, compensates for hardware drift, and achieves human-like performance across diverse experimental domains.
From complex optical to fragile quantum systems, this project aims at building the foundation for fully autonomous experimental labs of the future.
🧠 AI Framework
Built out of generative world models
State-of-the-art reinforcement learning methods.
Designed to adapt.
⚡ Features
Pre-trainable from existing datasets
Multi-domain adaptation, feature learning
Real-time control and resistant to drift and noise
