Ville Tuulos
Ville has been developing infrastructure for machine learning and AI for over two decades. He has worked as an ML researcher in academia and as a leader at a number of companies, including Netflix where he led the ML infrastructure team that created Metaflow, a popular open-source framework for ML infrastructure. He is the co-founder and CEO of Outerbounds, a company developing modern human-centric ML. He is also the author of a book, Effective Data Science Infrastructure, published by Manning.
Sessions
Over the past years, the compute landscape has got much more fragmented and heterogenous: GenAI needs access to various types of GPUs, sometimes leveraging vertical scalability, sometimes horizontal. The demand for CPU-based compute has got more diverse as well, as vertically scaling, high-performance data engines like Arrow and DuckDB have reduced need for inefficient approaches based on horizontal scaling. On top of this, the competition amongst clouds and specialized compute providers is getting more intense, motivated by customer demands for cost-efficiency.
Since its inception, Metaflow, which was originally open-sourced by Netflix in 2019, has been built to address diverse compute needs. Instead of proposing a new universal compute paradigm like Spark, which requires bespoke libraries, Metaflow integrates with various compute substrates and providers, including all major clouds. Recently, Metaflow gained support for large-scale distributed workloads, including distributed training on large GPU clusters.
In this talk, we give an overview of the changing landscape for compute and describe how open-source Metaflow allows Python developers leverage various compute platforms easily.