Founding QuantStack: Growing an Open Source Company to Global Impact (www.open-source-ward.com)

Sylvain Corlay, long-time Jupyter contributor, recounts founding QuantStack in 2016 to professionalize and fund sustained open-source work. What began as a one-person consultancy grown out of Bloomberg-era contributions now employs ~30 engineers and maintains critical infrastructure projects used across AI/ML: Jupyter/JupyterLab, conda-forge, Apache Arrow, and mamba (a high-performance conda installer). QuantStack is engineering-led, never took VC, and focuses on “infrastructure” problems—package management, tooling, and stable graphical interfaces—whose reach far exceeds the team size: JupyterLab sees ~500K PyPI downloads/day, Arrow is several times that, and an IBM study estimates tens of millions of users; Jupyter is even used at scale in French schools. For the AI/ML community this model matters because it shows how sustainable, multi-stakeholder maintenance enables reproducible tooling and performant data pipelines at scale. Corlay stresses governance: no single company should capture projects; collaboration—even with competitors—keeps shared infrastructure healthy (mamba being adopted into conda is a concrete example). His advice: join and maintain existing essential but under-resourced projects rather than launching new ones; that path both advances the ecosystem and creates service opportunities. QuantStack’s work and the upcoming PyData Paris / Arrow / Julia events underscore that foundational open-source engineering remains as strategic to ML progress as flashy trends like LLMs.