Tony is joined by Nathan Odle to talk about how he takes consumer hardware and puts together systems to do fun AI tasks. He has built systems using 7x NVIDIA 4090 and 7x Intel Arc A770. Come here about his experience building the systems and using the software to solve real problems and squeeze maximum performance out of GPUs you can buy yourself.

Guest:

Nathan Odle is an engineer, rally racer, and AI enthusiast who enjoys combining cutting-edge technology with practical applications. In the pursuit of backprop, Nathan recently built a custom AI server using seven Intel Arc A770 graphics cards, showcasing how consumer hardware can be used for advanced AI tasks. You can learn more about how to build your own on his website: http://mov-axbx.com

Nathan Odle

Resources:

Nathan's X

https://x.com/mov_axbx

https://www.intel.com/content/www/us/en/products/details/discrete-gpus/arc.html

Intel Extension for PyTorch

https://github.com/intel/intel-extension-for-pytorch

Quality of content and the web can be extremely challenging. Is the quality and civility of the content appropriate? Seekr is using AI to determine that for you as a business. They also are able to extend their learnings to help businesses build better LLMs for their own custom usages.

Stefanos Poulis, Chief of AI Research and Development, joins us to discuss how to build LLMs that measure civility or any other characteristic of a piece of content. He also talks about how they are exposing their capabilities to allow other businesses to do the same based on their own criteria

Stefanos Poulis

Rahul Nair

Learn More:

Seekr

https://www.seekr.com/

Intel® Liftoff for Startups

https://www.intel.com/content/www/us/en/developer/tools/oneapi/liftoff.htm

Intel® Tiber™ Developer Cloud

https://www.intel.com/content/www/us/en/developer/tools/devcloud/services.html

Intel® Gaudi® Al Accelerators

https://www.intel.com/content/www/us/en/products/details/processors/ai-accelerators/gaudi-overview.html

As AI applications become more prevalent, one of the biggest challenges is efficiently and cost effectively deploying your AI solution. For some it is as simple as following whatever pricing model a backend API SaaS from Google or OpenAI charges. But for others who are looking to deploy their own solutions, this can be challenging. By leveraging inefficiencies in the cloud compute market, NodeShift is making compute resources accessible to a wider range of businesses.

Guest:

Mihai Mărcuță is an Oxford University data science alumni with a professional background from Microsoft (Azure), Twitter (X), Epic Games (Fortnite) and Cisco. He co-founded NodeShift two years ago with Andrey Surkov.

Mihai Mărcuță

Resources:

Nodeshift

http://nodeshift.com

Intel Tiber Developer Cloud

https://www.intel.com/content/www/us/en/developer/tools/devcloud/services.html

Intel Tiber Developer Cloud Blog

https://community.intel.com/t5/Blogs/Tech-Innovation/Artificial-Intelligence-AI/AI-Cloud-for-Everyone-Nodeshift-Leverages-the-Intel-Tiber-Cloud/post/1598332

Intel Ignite

https://intelignite.com

Intel Liftoff for Startups

https://www.intel.com/content/www/us/en/developer/tools/oneapi/liftoff.html