Matt Squire is co-founder and chief technology officer of Fuzzy Labs, an AI development, implementation and consulting firm in Manchester, England, that helps companies run machine learning in production using open-source tools. He leads the firm’s technical work alongside its chief executive, Tom Stockton, and writes MLOps.WTF, a newsletter about the work of operating machine learning systems.
Making the Case for Open-Source MLOps
In late 2021 Fuzzy Labs reviewed the many tools available for running machine learning in production and published the results as a GitHub list, Awesome Open Source MLOps. He presented the work at MLOps London in January 2022 and laid out a starting set of tools: DVC to version data, Sacred to track experiments, ZenML to organize training into repeatable pipelines, Seldon Core to serve models behind web APIs, and Evidently to watch models for drift once they are live. “Open source gives us the freedom to pick and choose the best tools for every use-case, and have complete ownership over the solution,” he wrote afterwards.
In an Evidently community session that year he said the future lies in open source because “in almost all areas of tech, that’s already true,” and that a good tool should have a clear purpose and not try to do too many things. “Things should be Unix-like, in my view,” he said.
Bio-Inspired Computing and Side Projects
On his personal site he lists his interests as artificial intelligence, bio-inspired computing, internet-of-things devices and functional programming. The Fuzzy Labs team page says he spends spare time learning Mandarin and building his own internet-of-things projects.