AI for Science
Sadman Ahmed Shanto's work in scientific machine learning, autonomous experimentation, quantum-device inverse design, and AI-enabled research infrastructure.
I build machine-learning and agentic systems around scientific instruments, physics simulators, and engineering workflows. My work sits at the intersection of experimental physics, scientific software, and AI for science, with an emphasis on systems that produce testable physical outputs rather than text alone.
Research focus
Scientific machine learning for quantum-device design
I develop surrogate and inverse-design models that connect superconducting-circuit geometry to Hamiltonian parameters. This work builds on SQuADDS, an experimentally validated, open-source database and simulation workflow for superconducting quantum devices.
The research combines physics-aware feature engineering, neural networks, interpretable models, uncertainty-aware validation, and electromagnetic simulation. The goal is to shorten the path from a target device specification to a fabrication-ready candidate while keeping the result grounded in simulation and measurement.
Autonomous scientific workflows
I work on agent systems that can coordinate scientific tools across device design, simulation, experiment calibration, measurement, diagnostics, and analysis. My focus includes tool interfaces, workflow orchestration, evaluation, tracing, and safeguards that make autonomous scientific runs reproducible and inspectable.
Inference from experimental data
For superconducting-device experiments, I have used probabilistic sequence models to infer quasiparticle occupation from measurement trajectories. In earlier detector work, I developed machine-learning approaches for reconstructing missing events and improving three-dimensional muon-tomography pipelines.
Research software and computing infrastructure
My scientific-computing work spans Python and C++, high-performance computing, containerized services, APIs, data systems, electromagnetic solvers, and experiment-control stacks. I care about turning research prototypes into tools that other scientists can reproduce, validate, and extend.
Public evidence
| Area | Public artifact | What it demonstrates |
|---|---|---|
| Quantum-device design automation | SQuADDS paper, source code, and documentation | Experimentally validated simulation workflows, programmatic device generation, and open-source technical ownership |
| Scientific datasets | SQuADDS dataset | Reusable device-design and simulation data for computational research |
| Superconducting-circuit engineering | Design review and publication record | Breadth across design, simulation, measurement, and device physics |
| Scientific software | GitHub profile and project portfolio | Implemented research tools across quantum hardware, detector physics, simulation, and data analysis |
Ongoing or non-public work is described as such in my current CV; published papers and public repositories are linked separately so that readers can distinguish externally reviewable evidence from work in progress.
Relevant roles
This body of work is most directly relevant to scientific machine-learning, AI-for-science, research-software, autonomous-experimentation, quantum-device-design, and quantum-hardware R&D roles. The common thread is building computational systems that remain accountable to physical models, experiments, and engineering constraints.
Identity and profiles
For author disambiguation, publications may list me as Sadman Ahmed Shanto, Sadman Shanto, S. A. Shanto, or Shanto, S.A. My persistent identifier is ORCID 0000-0002-6325-4601. Additional records are available through Google Scholar, USC Physics, and GitHub.
Last updated: September 10, 2026.