Artwork

Content provided by S&P Global Market Intelligence and P Global Market Intelligence. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by S&P Global Market Intelligence and P Global Market Intelligence or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.
Player FM - Podcast App
Go offline with the Player FM app!

Data pipelines for AI

32:05
 
Share
 

Manage episode 454820584 series 2877784
Content provided by S&P Global Market Intelligence and P Global Market Intelligence. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by S&P Global Market Intelligence and P Global Market Intelligence or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.

Enterprises are wrestling with delivering data to fuel their AI efforts, hitting roadblocks around data security and privacy concerns and sifting through use cases and models to put it to work. Too many are making high-stake gambles feeding vast quantities of data into massive models. Jesse Robbins, one of the founders of Chef, a progenitor of the DevOps movement, a builder of the early Internet infrastructure and now partner at Heavybit, joins host Eric Hanselman to look at alternatives to the path that many are taking in pursuit of successful AI projects. In much the same way that DevOps patterns look to shift application development to more smaller, incremental changes with a pipeline that drives continuous improvement, AI projects can work with smaller models and localized datasets to manage risk and iterate faster. It’s a pattern that avoids concerns of pushing sensitive data to cloud-based offerings by working locally. Using smaller models reduces infrastructure costs and the need for vast quantities of GPU’s.

Larger models sizes and data sets create two problems – more computational power and supporting infrastructure is required and more data complicates data provenance, security and ownership issues. Starting smaller and expecting to iterate on the results locally can have multiple benefits. If the data being used never leaves the local confines, security concerns are constrained to local environments. Tools like the open source project Ollama can deliver a choice of models to fit a variety of use cases and infrastructure capacities. Just like DevOps patterns, starting small and iterating quickly can get further faster and with lower risk.

More S&P Global Content:

Credits:

  • Host/Author: Eric Hanselman
  • Guests: Jesse Robbins
  • Producer/Editor: Donovan Menard and Odesha Chan
  • Published With Assistance From: Sophie Carr, Feranmi Adeoshun, Kyra Smith

Other Resources:

  continue reading

102 episodes

Artwork

Data pipelines for AI

Next in Tech

13 subscribers

published

iconShare
 
Manage episode 454820584 series 2877784
Content provided by S&P Global Market Intelligence and P Global Market Intelligence. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by S&P Global Market Intelligence and P Global Market Intelligence or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://ppacc.player.fm/legal.

Enterprises are wrestling with delivering data to fuel their AI efforts, hitting roadblocks around data security and privacy concerns and sifting through use cases and models to put it to work. Too many are making high-stake gambles feeding vast quantities of data into massive models. Jesse Robbins, one of the founders of Chef, a progenitor of the DevOps movement, a builder of the early Internet infrastructure and now partner at Heavybit, joins host Eric Hanselman to look at alternatives to the path that many are taking in pursuit of successful AI projects. In much the same way that DevOps patterns look to shift application development to more smaller, incremental changes with a pipeline that drives continuous improvement, AI projects can work with smaller models and localized datasets to manage risk and iterate faster. It’s a pattern that avoids concerns of pushing sensitive data to cloud-based offerings by working locally. Using smaller models reduces infrastructure costs and the need for vast quantities of GPU’s.

Larger models sizes and data sets create two problems – more computational power and supporting infrastructure is required and more data complicates data provenance, security and ownership issues. Starting smaller and expecting to iterate on the results locally can have multiple benefits. If the data being used never leaves the local confines, security concerns are constrained to local environments. Tools like the open source project Ollama can deliver a choice of models to fit a variety of use cases and infrastructure capacities. Just like DevOps patterns, starting small and iterating quickly can get further faster and with lower risk.

More S&P Global Content:

Credits:

  • Host/Author: Eric Hanselman
  • Guests: Jesse Robbins
  • Producer/Editor: Donovan Menard and Odesha Chan
  • Published With Assistance From: Sophie Carr, Feranmi Adeoshun, Kyra Smith

Other Resources:

  continue reading

102 episodes

All episodes

×
 
Loading …

Welcome to Player FM!

Player FM is scanning the web for high-quality podcasts for you to enjoy right now. It's the best podcast app and works on Android, iPhone, and the web. Signup to sync subscriptions across devices.

 

Quick Reference Guide

Listen to this show while you explore
Play