Show Notes
Most AI pilots look impressive in a demo and never make it into production. Karthick Iyer has seen AI from the ground up and explains why, and what it actually takes to close that gap.
Karthick is Corporate Vice President and software leader in Intel's Client Compute and Physical AI Group. At the time of this conversation, he was Vice President of Systems Software at NVIDIA, where he led one of the teams building the software infrastructure that makes AI run at scale. Earlier in his career, he spent 18 years at Qualcomm, then worked at a startup and at Amazon.
For someone that close to the technology, his advice is refreshingly practical: start with your biggest pain point, and don't try to boil the ocean.
In This Episode
- What enterprise leaders can deploy today and see a return on, from defect detection on an assembly line to human-robot safety on the factory floor
- How mid-market companies without an AI team can compete, using AI computers that cost a few thousand dollars rather than hundreds of thousands
- Why the gap between a demo and a working deployment is bigger than most leaders expect
- Why forward deployed engineers matter more than coding agents for getting AI into production
- How to handle data silos, protect sensitive data and avoid vendor lock-in
- The five layers of AI, from power to application, walked through with a small shoe factory example
- How he uses an AI assistant to save two hours every morning
- Why AI should make your workforce more productive, not smaller
- Leadership lessons from working under Jensen Huang, including the "parking lot exercise"
- Work-life balance at the frontier of AI, and whether an MBA is still worth it
- His contrarian take on how much AI to use, and why humans stay in the loop
If you lead an operation and are trying to turn AI pilots into results, this episode gives you a grounded way to decide where to start and what to build.
Chapters
- 0:00 Introduction
- 2:21 What AI can deliver today: start with your biggest pain point
- 4:38 Competing with big players on a small budget
- 8:06 Misconceptions about AI and the demo-to-production gap
- 11:02 Why forward deployed engineers matter more than coding agents
- 14:56 Data silos and multi-agent orchestration
- 17:37 Data sovereignty and avoiding vendor lock-in
- 21:19 The five layers of AI
- 22:42 A small factory example, layer by layer
- 26:50 How he uses AI to save two hours every morning
- 28:08 The mental shift: AI as additional employees
- 29:57 Leadership lessons from working under Jensen Huang
- 34:14 Work-life balance at the frontier of AI
- 37:05 Is an MBA still worth it?
- 39:40 The contrarian take
- 42:02 Closing
Listen on Spotify and Apple Podcasts, and find more at leadingtransformationpodcast.com.
#EnterpriseAI #AIinManufacturing #digitaltransformation