A Year in Review: Quansight’s Contributions to PyTorch in 2023 (& Early 2024)

PyTorch logo

2023 will be remembered as the year when AI and LLMs took the world by storm. PyTorch took center stage during this revolution due to the rise of torch.compile. The combination of having a fully flexible eager execution model, paired with a compiler with a rather flexible tracer that is able to understand complex Python programs semantically, has certainly been one of the core components fueling these advances.

Make Your AI Vision a Reality with Quansight’s AI Engineering Consulting

Artificial Intelligence (AI) is reshaping industries worldwide, driving innovation and pioneering new possibilities. The journey from AI research to practical, enterprise-ready solutions can be challenging, and this is where Quansight’s AI Engineering consulting services come into play. With our deep expertise in open source scientific computing

Ragna in Action: Building AI Document Interrogation Apps with Open Source Tools

The image shows the logo for Ragna. It features a stylized Viking ship with a dragon figurehead and three round shields along the side. The sail is depicted as a reddish-orange sheet of paper with white lines representing text. Below the ship, the word "RAGNA" is written in a modern, bold font. This logo symbolizes Ragna’s role in navigating and harnessing the power of data and AI technology, much like a Viking ship exploring and conquering new territories. The words, Ragna In Action appear below.

A look at recent presentations on AI, RAG, and Ragna by Quansight’s staff. “For playing around, Generative AI is definitely cool. For asking serious questions about documents, which you’re going to base your business on, I think you still need to do your research and work out how to put some guard rails on it.” […]

Building a GPU CI Service for conda-forge

Logos of MetroStar Systems, conda-forge, and Quansight, representing their collaboration.

The recent revision of the array API standard marks another major milestone in the collective effort to achieve array interoperability across the Python data ecosystem.

A recent post by Quansight Lab’s Athan Reines on the Data APIs blog shares updates on the consortium’s progress and plans for the future.

Introducing Lightweight Versions of GDAL and PDAL

Geocoding is the practice of taking in an address and assigning a latitude-longitude coordinate to it. Doing so for millions of rows can be an expensive and slow process, as it typically relies on paid API services. Learn how we saved our client time and money by leveraging open source tools and datasets for their geocoding needs.

Reverse-Geocoding in AWS Lambda: Save Time and Money Using Polars Plugins

A watercolor illustration of a polar bear plugging an electric plug into a wall socket, symbolizing the connection and power of using Polars for geocoding and reverse-geocoding tasks.

Geocoding is the practice of taking in an address and assigning a latitude-longitude coordinate to it. Doing so for millions of rows can be an expensive and slow process, as it typically relies on paid API services. Learn how we saved our client time and money by leveraging open source tools and datasets for their geocoding needs.

Community-First Open Source: An Action Plan!

Quansight CTO Dharhas Pothina is here to discuss an action plan for achieving community-driven open source. What is Community-first Open Source? Most open source software (OSS) projects our readers work on are community-first or community-driven. Today, we will explore the transition from company-backed to community-driven open source, define the two, and explain why we are […]

Taming the Toxic Python Environment on your Laptop

Local Reproducible Open Source Environments with conda-store Are you tired of running into problems every time you install a new Python package, messing up your existing code in Jupyter notebooks? Have you ever told a colleague, “Well, it works on my computer,” only to find out the issue is specific to their environment? If so, […]

PyTorch Progress as a General-Purpose Array-Oriented Computational Backend

This post was created from Travis Oliphant’s keynote at PyTorch Conference 2023. The full recording is linked below. In recent years, the landscape of scientific and engineering computing has undergone a transformative shift, propelled by advancements in array-oriented programming. At the heart of this evolution lies the SciPy, PyData, and SciPyData ecosystems, which have long […]