About

Grey Haven Solutions offers consulting and software development services for computational research. With peer reviewed papers across the fields of physics, materials engineering, and computer science, owner-operator Liam Huber brings nearly two decades of experience in computational science and research computing.

Services

• Research consulting – We offer consulting on computational science projects with a specialization in materials science and atomistic simulations. In 2018, Liam wrote the seminal paper on using machine-learned surrogate models for predicting solute-defect interaction energies in metals.
• Software development – Whether it’s developing a custom software tool or computational model to tackle your problem, or converting a student’s unreadable thesis code into a reusable library for future use by your entire lab, we can work with you to solve your computational and software needs.
• Training – Leverage over a decade of experience in speaking, teaching, and leading small development teams in a workshop for your team. Topics cover both hard and soft skills including: architecture patterns in python for research software, version control with git, best practices for asynchronous communication (i.e. how to write a good pull request, how to review your colleagues’ code, etc.), tools for distributing and sharing your code, and how to effectively present your code and results.

Showcase

The most recent Grey Havens Solutions project is pyiron_workflow, commissioned by the Max-Planck-Institute für Nachhaltige Materialien. pyiron_workflow is a workflow management tool for bringing rigor, reproducibility, scalability to the world of academic python. Using a graph-and-node based computation paradigm, pyiron_workflow abstracts many technical details to allow scientists to write their workflows with simple decorated python functions. While computational power can be scaled by sending individual nodes in the graph to parallel processes, workflow complexity can be scaled by composing macros – not just for DAGs, but for cyclic graphs too. Formalizing the flow of data into a graph paradigm gives reliability to research code, and even allows python to be used in a strongly-typed way.

What users are saying about pyiron_workflow:

Activity

Research consulting

On a variety of topics related to atomistic simulation of materials, including:

• Building machine-learned surrogate models for calculating energetic properties.
• Mean-field models to access free energies without sampling
• Generating training sets for transferable machine-learned interatomic potentials

Research Software DevOps

In pyiron/actions, we centralized and expanded the GitHub actions and workflows for the pyiron project; e.g. robust caching and retrieval of Conda environments, automated range-bounds for semantically versioned dependencies calculated at deployment time, etc.

NumFocus Small Development Grant – Ontological typing

Grey Haven Solutions previously received a 2022 NumFocus Small Development Grant to develop a system of ontological typing. By encoding workflows as a graph and defining ontological types for the data moving through these workflows, we can leverage ontological reasoning software to dynamically narrow types based on how data is used downstream or what the upstream requirements are in the workflow. This tool is available via Conda-Forge, and applications for guided workflow design can be seen in the ironflow, a prototype GUI built for the pyiron project.

Contact

If you think Grey Haven Solutions may be of assistance to you, please write to info@greyhavensolutions.com to discuss your project!