HighRes® Positions Cellario OS™ as the Open Execution Layer for AI Scientists, Beginning with NVIDIA BioNeMo Agent Toolkit

Cellario OS™ is the open, instrument-agnostic execution layer that turns experiments from any AI Scientist into validated, provenance-rich laboratory data, completing the loop between AI-driven hypothesis and real-world result.

BEVERLY, Mass.--(BUSINESS WIRE)--HighRes® today announced that its Cellario OS™ lab orchestration platform serves as the open execution layer for AI Scientists, both autonomous agents and AI-enabled researchers, giving any AI Scientist a reliable path from a computational hypothesis to a physically executed experiment. Cellario OS integrates with NVIDIA BioNeMo Agent Toolkit, open models and skills that accelerate the rate of hypothesis generation, and is built to connect to the broader, fast-growing ecosystem of AI Scientist platforms. Whichever AI Scientist a customer chooses, HighRes provides the execution layer that designs, runs, and validates experiments on real instruments and returns structured, trustworthy data the AI Scientist can act on.

Agentic life sciences workflows can generate and evaluate experiments at machine speed, but a protocol remains a hypothesis until something runs it correctly and returns data that can be trusted. As AI Scientists accelerate the rate of hypothesis generation, the constraint shifts to execution, testing more ideas with greater speed, efficiency, cost-effectiveness, reproducibility, and automation. Cellario OS and the Cellario Lab Assistant™ orchestrate instruments, workflows, and data into reliable, repeatable physical execution. They sit between the AI Scientist, which forms the hypothesis and designs the experiment, and the bench, where instruments and samples do the work.

A central part of the approach is capability discovery. Rather than designing against a generic template, an AI Scientist can first query the lab through the Cellario Lab Assistant for the instruments, assays, methods, labware, volumes, throughput, and constraints available on the floor. The lab returns a machine-readable capability model, so experiments are designed against real-world conditions rather than assumptions.

Within this framework, Cellario OS translates scientific intent into an executable workflow bound to a lab’s specific instruments and validated methods. Generated protocols are simulated in Cellario before any hardware moves, with scientists in the loop to optimize experimental design. The lab layout can be visualized and rendered in a physically accurate way using NVIDIA Omniverse libraries integrated into Cellario Lab Designer™. Cellario OS then orchestrates instruments and robotics deterministically with full traceability of every step, sample, and device. Using physical AI tools like NVIDIA Cosmos, lab instruments and workflow anomalies are detected in real time so the system can reason on autonomous recovery actions or error escalations. Results return to the AI Scientist as quality-controlled, contextualized, provenance-rich data that includes what ran, on which instrument, under what conditions, and with what deviations, thereby closing the loop.

The execution layer is designed to be open and instrument-agnostic. Cellario OS sits beneath the AI Scientist of the customer's choosing, whether that’s an organization's own in-house agents or an R&D platform such as Benchling, where HighRes already serves as the execution layer through an established partnership, integrating with, rather than replacing, existing systems, so a customer's models, data, and intellectual property remain their own. HighRes can bring NVIDIA BioNeMo models directly into the Cellario Lab Assistant; at the same time, the company expects the industry to converge on specialized AI Scientists interfacing with Cellario OS as the shared capability and execution layer that designs experiments, runs them on the bench, and returns validated results.

“An AI Scientist can design thousands of experiments, but none of them are real until the lab runs one and returns data you can trust,” said Ira Hoffman, CEO of HighRes. “We're building Cellario OS to be the open execution layer for every AI Scientist, starting with NVIDIA BioNeMo to accelerate hypothesis generation, and extending to the in-house and commercial agents our customers are adopting. The AI Scientist designs; Cellario OS runs the experiment and returns results you can trust.”

NVIDIA today introduced NVIDIA BioNeMo Agent Toolkit at the BIO International Convention, providing domain-specific tools and skills that let AI agents synthesize scientific knowledge, call models, evaluate results, and execute next actions across biology, chemistry, genomics, and drug discovery. HighRes extends those agent-callable skills into the physical lab, and applies the same open execution layer to any AI Scientist a customer chooses, providing the path that connects computational discovery to laboratory results.

For more on NVIDIA BioNeMo Agent Toolkit, read the full NVIDIA announcement here.

ABOUT HIGHRES

HighRes partners with life science organizations to implement intelligent data and lab automation workflows that empower every team member to plan and execute efficient and reproducible science. Learn more at highres.com.