Pasqal is addressing a core challenge in quantum computing: translating theoretical ideas into working experiments. While cloud access has broadened reach to quantum processors, the company notes that running a useful experiment still demands considerable technical coordination across physics, software, emulation and cloud execution. Their new agentic workflow aims to automate these implementation steps, supporting the process. Pasqal emphasizes the workflow is designed to keep researchers responsible for “what to test, how to validate the experiment, and what the results mean.”
Agentic Workflow: From Scientific Objective to Experiment
The agentic workflow automates the conversion of a scientific objective into a runnable quantum experiment by first extracting relevant physical parameters, constraints and hardware assumptions from initial ideas or published papers into a structured file named experiment_spec. json. This structured file is a reference input for all subsequent steps, allowing researchers to review, modify, or approve the proposed experiment before implementation proceeds. Pasqal specifically designed this initial stage to ensure alignment between the theoretical intent and the eventual hardware execution, a critical step often requiring significant manual effort.
The system then moves beyond specification, testing the feasibility of translating theoretical work into a practical quantum program. One demonstration revisits a 2019 Rydberg-array experiment focused on density-wave ordering, assessing whether the original findings can be replicated on current hardware.
Another case examines a patent, determining if the described protocol is implementable on a Pasqal QPU, and demonstrating the workflow’s capacity to automate successful conversions while identifying areas still requiring expert oversight, the company says. The team emphasizes that achieving runnable code and plausible results is insufficient; researchers must still verify that the experiment and its observables address the original scientific question.
The workflow is not intended to replace scientific judgment, but rather to alleviate the burden of tedious implementation details, according to Pasqal. “I have to choose the best configuration for a problem with 50 discrete variables.” This division of labor is deliberate, ensuring researchers retain control over the scientific process while benefiting from automated assistance with the technical complexities of neutral-atom systems.
QPU submission is handled through Pasqal Cloud, integrating the automated processes with a specific hardware platform. The system also supports scenarios where a researcher might ask, “I want to know whether a square array of about 25 atoms orders antiferromagnetically when I ramp the detuning through the transition.
Set it up so I can emulate it locally first.” or even, “I have access to a neutral-atom machine and I would like to do an experiment on thermalization. I don’t know what to measure.”, demonstrating its adaptability to different stages of experimental design and analysis. The Neutral-Atom Agentic Toolkit is now available for researchers interested in testing the workflow with their own published protocols or research ideas, with detailed analysis of the underlying methodology available on arXiv.
Pulser Sequence Generation & Validation with Emulation
Pasqal’s agentic workflow coordinates multiple steps, physics, software, emulation and cloud execution, previously handled separately by researchers, streamlining the process of translating a scientific objective into a functioning quantum experiment. The system uses existing Pasqal software, including Pulser for pulse-sequence design, emu-mps for emulation and the Pasqal Cloud SDK for device access and execution, integrating them into a cohesive automated process. This integration allows for the conversion of a high-level experiment specification into an executable Pulser sequence, specifically tailored for neutral-atom experiments by defining the atom register and laser controls.
Before submission to quantum processing units, the workflow validates candidate sequences through emulation, comparing ideal behavior with simulations that incorporate device-aware noise. The skills validate-emu and noise-emulate estimate whether a target observable will remain visible under realistic conditions, identifying potential hardware limitations before valuable QPU time is used.
Researchers recently tested this validation process by deliberately tasking the agent with reproducing a phase known to be infeasible on their target QPU, demonstrating the system’s ability to correctly identify hardware constraints and redirect implementation toward a feasible regime. The agent then generated the corresponding sequence, validated it in emulation, and supported execution on Pasqal QPUs, showing a complete cycle from specification to data acquisition.
However, the system’s limitations also surfaced during testing, revealing the continued need for expert scientific judgment. The agent initially focused on observables easy to compute but insufficient to certify the target phase, producing plausible data from a scientifically incomplete diagnostic. A domain expert then redirected the analysis toward the appropriate order parameter and finite-size scaling strategy, highlighting that the workflow is designed to alleviate implementation burdens, not replace the core scientific reasoning.
This approach allows the workflow to function as a tool for connecting scientific demand with future hardware development, providing a measurable scan of current hardware reach. The agentic workflow’s skills are now available as plugins for coding environments like Claude Code, Cursor and Codex, further lowering the implementation barrier for quantum experimentation.
Pasqal’s system supports the entire process, from protocol extraction and sequence generation to emulation, QPU submission through Pasqal Cloud and subsequent analysis, offering a comprehensive solution for researchers seeking to use quantum hardware effectively. The research examples demonstrate how the workflow can coordinate these steps, enabling more efficient exploration of quantum phenomena.
Pasqal Cloud SDK for QPU Submission & Data Analysis
Pasqal’s system now coordinates the entire quantum experiment lifecycle, from initial protocol extraction to data analysis, using the Pasqal Cloud SDK for QPU submission and device access. This integration extends beyond simply providing cloud access. It actively manages the complex interplay between physics, software, emulation and hardware execution, a historically fragmented process. The agentic workflow’s ability to handle multi-round experiments, as demonstrated by a project based on a Pasqal patent for Rydberg-based graph coloring, highlights a significant advancement in complexity, the firm reports.
Unlike simpler QPU submissions, this algorithm required iterative execution, with each round dependent on the results of the previous one, demanding a system capable of managing dependencies and data flow across multiple hardware runs. This capability is about orchestrating a series of interconnected operations, a task previously requiring substantial manual effort.
The system also incorporates calibration information from the device’s current operating point, ensuring experiments are tailored to the specific characteristics of the hardware. Recent testing revealed that successful execution does not guarantee scientific validity, a distinction often overlooked in early-stage quantum computing. EU-based researchers benefit from free access to the Pasqal QPU hosted at TGCC-GENCI, facilitated by the apply-genci-tgcc-cea skill within the workflow. Once QPU jobs are completed, the harvest-and-analyze skill retrieves measurement data and converts raw bitstrings into the observables defined in the experiment specification.
This automated data processing streamlines the analysis phase, allowing researchers to focus on interpreting the results rather than managing data formats and conversions. The ability to scan and assess hardware reach also connects current scientific demands with the development of future hardware capabilities.
Workflow Testing: Reproducing Rydberg-Array Experiments
Submitting experiments to quantum processing units (QPUs) through Pasqal Cloud now incorporates device calibration data, ensuring experiments account for the current operating conditions of the hardware. This automated inclusion of calibration information streamlines the process, moving beyond simple cloud access to a more integrated system for experiment execution. Recent testing of the agentic workflow involved reproducing a 2019 Rydberg-array experiment investigating density-wave ordering, a case designed to stress the system’s capabilities, Pasqal reports.
While the workflow successfully generated a runnable experiment, its initial approach revealed a critical distinction between execution and scientific validity. The agent first prioritized easily computable observables, producing plausible data that nonetheless failed to adequately confirm the intended target phase.
This outcome highlights that automated execution does not guarantee meaningful scientific insight, and expert oversight remains essential for interpreting results. The system’s limitations became apparent when the agent’s initial analysis proved incomplete, prompting intervention from a domain expert. This expert redirected the workflow toward a more appropriate order parameter and a refined finite-size scaling strategy, demonstrating the need for human guidance in complex quantum investigations.
The team reports that this case, alongside two others, one recreating a theory paper and another implementing a patent, demonstrates both the successes and necessary limitations of the automated workflow. The workflow’s ability to generate experiments for accessible regimes, even when initial attempts are flawed, suggests a valuable tool for accelerating quantum research.
These tests also revealed that the workflow can successfully automate certain steps, such as sequence generation and validation, but requires expert review for scientific diagnostics. “The more interesting result was not that the workflow produced a runnable experiment, but where it failed,” the researchers noted, emphasizing the importance of understanding the boundaries of automation in quantum computing. The team’s work demonstrates a shift toward a system where automation handles implementation details, while researchers retain responsibility for defining the scientific goals, validating the experiments and interpreting the results.
Identifying Feasible Regimes for Quantum Magnetism
The agentic workflow successfully identified regimes inaccessible to current hardware during tests recreating a 2019 Rydberg-array experiment on density-wave ordering, demonstrating a capability beyond simply generating runnable code. The agent correctly flagged the limitation, then autonomously adjusted the implementation parameters to a viable regime, generating and validating the necessary experimental sequence for execution on Pasqal QPUs. This initial test extended beyond simple protocol translation; the resulting density-density correlations accurately mirrored expected physical behavior, confirming the workflow’s ability to bridge the gap between theoretical design and physical realization.
A second, more complex test involved a theory paper detailing frustrated quantum magnetism in a triangular Rydberg array, where the agent’s role was to determine which proposed phases of matter were realistically achievable given hardware constraints. The system compared the requirements of each phase against the available resources, categorizing them as feasible, marginal, or infeasible, a critical step in streamlining experimental design. The workflow’s ability to assess feasibility is particularly valuable given the challenges of translating theoretical proposals into concrete hardware experiments.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
