AI news ·
Agentic AI AutoLabs helps scientists instruct laboratory robots faster
PNNL researchers built AutoLabs, an AI that translates scientific goals into robot instructions. It cuts lab setup from weeks to minutes and boosts throughput tenfold.
Researchers at the Pacific Northwest National Laboratory developed an agentic AI system that translates experimental goals into direct instructions for laboratory robots, cutting experiment setup times from weeks to minutes. The system allows scientists to bypass the lengthy engineering coordination typically required for autonomous lab equipment, potentially increasing experiment throughput by up to tenfold.
Bridging the knowledge gap
Autonomous laboratory equipment requires specialized programming. Scientists understand the experimental goals, while systems engineers understand the mechanical limits of the hardware. Translating a research question into machine code usually demands weeks of collaboration between the two disciplines.
To solve this, PNNL data scientist Gihan Panapitiya and his team built AutoLabs. The system uses large language models to bridge this gap, functioning as a practical application of AI Agents & Automation in a physical laboratory setting. AutoLabs operates on an OpenAI model and uses multiple specialized sub-agents managed by a single supervisor agent to interpret user requests.
Testing the translation layer
The team tested AutoLabs by feeding it five experimental tasks of increasing complexity for Big Kahuna, an automated robot built by Unchained Labs that studies battery materials. The simplest task required mixing specific ratios of naphthalene and methanol. The most complex involved multi-chemical reactions with strict temperature controls, specific stirring speeds, and liquid transfers between vials.
AutoLabs successfully generated the correct hardware files for every test. "With AutoLabs, human experts can learn to use Big Kahuna quickly and guide the overall experimental strategy while the AI agent manages the granular implementation and validation," said Heather Job, a systems engineer at PNNL.
Adapting to other laboratory systems
The software is available on GitHub and designed to adapt to other autonomous laboratory systems. The development team is currently adding memory to the system, allowing it to conduct literature reviews and learn from past experiments over time.
"This iteration of AutoLabs was developed to generate hardware files for Big Kahuna, but its experiment design capabilities extend beyond this one robot," Panapitiya said. The goal is to create a self-correcting partner for scientific discovery rather than a replacement for human researchers.
Why this matters for science and research professionals
Laboratory automation has historically required dedicated engineering staff to operate complex machinery. Tools like AutoLabs shift this dynamic, giving researchers direct control over robotic workflows without needing deep programming expertise. As institutions explore AI for Science & Research, delegating routine machine translation to agentic systems will allow scientists to focus entirely on experimental design and data analysis.