Designing a Data Science Workstation: Part I

Part I in a multi-part series on designing and building my own personal Data Science Workstation. From motivations of design based on my decade's worth of work in the ML field to the hardware and software considerations that make up data science performance, this series is made to help those in the data science and AI space to learn from my experiences and find a system design that works for them.

#Datascience #MachineLearning #AI #DataScienceWorkstation

Full Disclosure: I work for Intel

00:00 Intro
00:37 Overview and Problem Statement
03:02 NUMA and Network NUMA
03:29 Bias by Model Selection
05:04 Building the System
05:15 Optane Persistent Memory
06:24 CPU selection
07:24 Motherboard Selection
07:49 Memory Configuration
08:29 Chassis Config
08:35 Display Adapter
09:24 Initial system test

EU regulations on high-risk AI: https://ec.europa.eu/info/strategy/pr...

SciPy talk on Oversubscription:    • Addressing Multithreading & Multiproc...  

Looking at oversubscription in AI frameworks:   / choosing-the-right-intel-workstation-proce...