In Episode 3 of AI Factory Insider, host Kaushik Shirhatti (VP, AI Factory, NVIDIA) is joined by Anne Hecht (Sr. Director of Product Marketing, NVIDIA) to discuss autonomous agents that continuously learn and compound value, purpose-built AI factory infrastructure, and real-world use cases.
What we cover
0:00 — Introduction
1:43 — Meet our guest: Anne Hecht
3:35 — From gen AI to agentic AI: what actually changed
5:20 — What broke the dam: enterprises seeing real ROI from their AI projects
6:50 — Inside NVIDIA: 4 trillion tokens a day
7:17 — Three words to describe AI agents: persistent, autonomous, self-evolving
11:05 — The CFO view: tokens, marginal cost, and where AI runs
14:16 — The IT & CISO view: manage agents like you manage employees
16:28 — "Your agent is only as good as its connections"
17:59 — Shadow AI is the new shadow IT
18:51 — What's in it for users: privacy, confidentiality, and control
19:58 — Easy is hard: designing a seamless agent experience
21:02 — The GTC slide: Agent = harness, models, and tools
23:06 — Agent skills & why fine-tuned open models cut cost
27:24 — Securing agents: the Secure Agent Workspace & OpenShell
29:23 — Confidential computing: protecting models & data in use
30:36 — Product or reference design? What's what
32:52 — Frontier vs. open models: it's an "and," not an "or"
34:35 — The Linux parallel: open source as the OS of compute
36:51 — Nemotron adoption & post-training
37:39 — Jensen Huang on X & the Open Secure AI Alliance
40:14 — Wrap-up
Resources
Secure Agent Workspace Reference Design
NVIDIA OpenShell
NVIDIA AI Factory Validated Design Guide
AI Factories for Self-Evolving Agents (Kaushik & Anne's GTC Taipei session)
Deploying Proprietary Models Securely with NVIDIA Confidential Computing on Self-Hosted VMs
Deploying Proprietary Models Securely with NVIDIA Confidential Computing on Self-Hosted Kubernetes
Building an Analysis AI Agent for Industrial Alarm Management with NVIDIA Nemotron
What we cover
0:00 — Introduction
1:43 — Meet our guest: Anne Hecht
3:35 — From gen AI to agentic AI: what actually changed
5:20 — What broke the dam: enterprises seeing real ROI from their AI projects
6:50 — Inside NVIDIA: 4 trillion tokens a day
7:17 — Three words to describe AI agents: persistent, autonomous, self-evolving
11:05 — The CFO view: tokens, marginal cost, and where AI runs
14:16 — The IT & CISO view: manage agents like you manage employees
16:28 — "Your agent is only as good as its connections"
17:59 — Shadow AI is the new shadow IT
18:51 — What's in it for users: privacy, confidentiality, and control
19:58 — Easy is hard: designing a seamless agent experience
21:02 — The GTC slide: Agent = harness, models, and tools
23:06 — Agent skills & why fine-tuned open models cut cost
27:24 — Securing agents: the Secure Agent Workspace & OpenShell
29:23 — Confidential computing: protecting models & data in use
30:36 — Product or reference design? What's what
32:52 — Frontier vs. open models: it's an "and," not an "or"
34:35 — The Linux parallel: open source as the OS of compute
36:51 — Nemotron adoption & post-training
37:39 — Jensen Huang on X & the Open Secure AI Alliance
40:14 — Wrap-up
Resources
Secure Agent Workspace Reference Design
NVIDIA OpenShell
NVIDIA AI Factory Validated Design Guide
AI Factories for Self-Evolving Agents (Kaushik & Anne's GTC Taipei session)
Deploying Proprietary Models Securely with NVIDIA Confidential Computing on Self-Hosted VMs
Deploying Proprietary Models Securely with NVIDIA Confidential Computing on Self-Hosted Kubernetes
Building an Analysis AI Agent for Industrial Alarm Management with NVIDIA Nemotron
This website uses cookies
Comments 0
Be the first to comment!