Innovation moves fast but production environments do not. Often research groups or individuals install everything with no controls, giving them the flexibility in the short term, but once they move it into production, everything fails.
Packaging—the process of preparing and distributing software libraries or applications so they can be easily installed and used—is crucial for making software accessible, manageable, and shareable across the Python community and beyond.
Finding the right partner who understands these challenges is key to ensuring your packaging and environment management processes are reliable and scalable.
At Quansight, our open source experts and decision-makers include steering committee members and maintainers of core projects. Additionally, we created conda-store, an open source tool focused on collaborative data science environments.
We have Steering Council Members & Contributors on our team. Conda provides package, dependency, and environment management for any language.
We employ Core Team Members and Contributors. Conda-forge is a community-led collection of recipes, build infrastructure, and distributions for the conda package manager.
Quansight is the creator of this resource, which addresses the unique challenges of packaging Python projects and provides key insights and references.
As maintainers, we leverage the Meson build system to offer robust build backends for Python packages. Our team members successfully transitioned several key projects to use this system.
We help maintain critical packages under the Python Packaging Authority (PyPA), contribute to the authorship of Python Enhancement Proposals (PEPs), and actively drive the evolution of Python packaging standards.
Our team incubated this tool to simplify the creation, management, and collaboration of data science environments.
“Point72 and Cubist are committed to open source and to sponsoring organizations such as PyData and the Python Software Foundation. We are excited about the opportunities our partnership with Quansight may provide to solve packaging problems strategically and sustainably both for our own research teams and for conda-forge users generally.”
Reproducibility cannot be an afterthought. Achieving it requires clearly defined research and production environments that remain adaptable across various contexts. This proactive integration into the development process addresses challenges such as rapidly evolving libraries and complex workflows.
By setting clear objectives and identifying potential risks, we help practitioners create robust systems that enhance the reliability and credibility of their work.
Rapid library evolution can introduce breaking changes that compromise reproducibility.
Practitioners often work in varied environments, making it difficult to ensure consistent results across different setups
IT departments may enforce strict control over environments, conflicting with the flexibility needed by data scientists.
Existing workflows often do not support reproducibility, making it challenging to replicate results or share work.
