NVIDIA NVIDIA
September 29, 2026

Vertically Integrated, Horizontally Open | AI Factory Insider Ep 5

How can an AI platform be optimized from hardware to software while staying open to developers and partners at every layer?

In Episode 5 of AI Factory Insider, host Kaushik Shirhatti (VP, AI Factory, NVIDIA) is joined by Pradeep Gupta (VP, Industries Solutions Architecture, NVIDIA) and Stephen Jones (CUDA Architect, NVIDIA) to discuss how vertical integration works with an open ecosystem, how CUDA-X libraries serve specialized industry workloads, and how agent skills are changing software development.

What we cover
0:00 — Introduction
1:08 — Meet Pradeep Gupta and Stephen Jones
1:28 — What does "vertically integrated, horizontally open" mean?
3:25 — One platform, many industries
4:30 — AI across healthcare, retail, and finance
6:35 — What stays common across industries
8:55 — CUDA: the link from apps to GPUs
10:29 — CUDA's early days and the first debugger
13:24 — The spreadsheet behind CUDA profiling
14:24 — CUDA vs. CUDA-X libraries
16:08 — Identifying the right CUDA-X library for your use case
17:42 — How is CUDA open if it runs on NVIDIA GPUs?
21:33 — CUDA-X in finance, biopharma, and smart cities
22:50 — How CUDA libraries build on each other
25:18 — CUDA and GPU hardware co-design
28:05 — Reuse the core; specialize when needed
29:15 — Agents need tools and skills
31:01 — CUDA-X libraries as agent skills
32:59 — How AI agents change the way developer work
37:15 — Wrap-up

Resources
CUDA: New Features and Beyond
NVIDIA Enterprise AI Factory Design Guide White Paper
CUDA-X Libraries
CUDA Toolkit
NVIDIA AI Factories
214

Comments 0

Replying to
5000 max

Be the first to comment!

This website uses cookies
This website uses cookies.
To maintain our services, we use cookies. They allow us to personalize ads, analyze our traffic, and provide social media features.

Customize your preferences

Select the cookies you accept: