Podcast Summary:
Reshaping Workflows with Dell Pro Precision and NVIDIA RTX PRO GPUs
Episode: How AI and Accelerated Computing Reshape Manufacturing
Host: Logan Lawler
Guest: Himanshu Iyer, NVIDIA Industry Marketing Lead (Manufacturing)
Date: April 9, 2026
Overview
This episode explores how AI and accelerated computing are transforming the manufacturing sector, focusing on the integration of Dell Pro Precision workstations with NVIDIA RTX Pro GPUs. Host Logan Lawler and guest Himanshu Iyer discuss the evolution of manufacturing workflows, the shift from CPU to GPU computing, the rise of AI (including agentic and physical AI), challenges in data integration, and practical advice for organizations adopting these technologies.
Key Discussion Points & Insights
1. Complexity & Multi-Step Nature of Manufacturing
[02:59]
- Manufacturing is a highly complex and multi-stage process: from requirements gathering, R&D, conceptual and detailed design, simulation, production, assembly, to documentation.
- "Manufacturing is one of the oldest industries... These designers, engineers, manufacturing experts...have come up with processes to improve efficiency, but there are still plenty of opportunities to harness the power of accelerated computing... to revolutionize how products are designed and made." — Himanshu Iyer [03:42]
- Dell and NVIDIA provide the hardware/software stack to inject efficiencies at every stage.
2. Evolution from CPU to GPU Computing in Simulations
[05:57]
- Engineering simulations (for stress, temperature, vibration testing, etc.) were traditionally performed on CPUs, often requiring servers or data centers, taking hours to weeks.
- Shift to running simulation solvers on GPUs has made the process 10x–100x faster due to parallelization.
- “A simulation that would have taken hours or days is 10x, 20x, or even 50–100x faster, depending on the application.” — Himanshu Iyer [07:57]
- Modern workstations (e.g. Dell Pro Precision with multiple NVIDIA GPUs) can now run complex simulations at desktop, reducing cloud/data center dependence and cost.
3. Data, Silos, and AI in Manufacturing
[12:29]
- Manufacturing data is vast, often siloed across PLM, ERP, MES, and IoT systems, sometimes spanning decades.
- "There needs to be some data governance policy ... so they can then build intelligence out of all of that data rather than having to work with each system individually." — Himanshu Iyer [13:20]
- Emergence of the 'digital thread'—a continuous data flow connecting all stages and teams.
- AI (including agentic AI) can unify these silos, automate tasks, run multi-stage workflows, and empower both design and simulation through actionable consolidation.
4. Rise of Agentic AI
[14:35]
- From perceptual and generative AI, industry is now moving to agentic AI: autonomous agents capable of executing bounded workflows and iterative experimentation.
- "Agentic AIs are assisting designers and engineers at the stage they are working on for the manufacturing process." — Himanshu Iyer [14:36]
- Key benefit: local, rapid iteration directly on workstation resources, without waiting for remote compute.
5. Scaling AI: From Proof of Concept to Organization-Wide
[17:02]
- AI in manufacturing is transitioning from PoC to large-scale, mission-critical deployment, particularly in quality control and inspection.
- IoT devices provide streams of sensor data, which AI agents monitor in real time, sending alerts and predicting anomalies.
- Scaling requires:
- Unified teams and clear KPIs
- Management support
- Robust, scalable compute infrastructure (favoring flexible, modular GPU setups)
- Team training and skill development
6. Center of Excellence vs. Decentralized Adoption
[21:39]
- Trade-off: Experimentation is necessary, but too much decentralization leads back to siloed approaches.
- "Forming a core team, a center of excellence, will make sure that this siloed approach doesn't take hold...that digital thread is connecting all of these different models." — Himanshu Iyer [23:20]
- Center of excellence provides data and model governance, ensures organization-wide alignment, and maximizes ROI and data reusability, albeit with slower initial rollout than decentralized methods.
7. The Next Frontier: Physical AI
[25:47]
- AI's journey: Perceptive AI → Generative AI → Agentic AI → Physical AI (physics-aware models).
- Physical AI: models understand and simulate laws of physics (e.g., gravity, collision, friction) and are used to train robots, autonomous vehicles, and digital twins with synthetic, physically accurate data before real-world deployment.
- "Where we are right now, LLMs don’t necessarily fully understand the laws of physics...That is where journey takes us: new models will be physically aware." — Himanshu Iyer [26:40]
- NVIDIA’s Cosmos foundation model is an example—enabling physics-accurate digital simulations for robotics and assembly line automation.
8. Benefits of Physical AI in Manufacturing
[33:41]
- Robotics is a prime benefactor—addressing skills/talent gaps amidst workforce retirements.
- Physical AI models drastically reduce time, risk, and cost by simulating millions of hours and scenarios (weather, failures, variations) virtually.
- “Doing all of this testing or simulations physically is impossible...doing all of this virtually in physics AI models is the only way we can move forward.” — Himanshu Iyer [34:38]
9. Practical Advice: Five Keys to Preparing for AI in Manufacturing
[36:54]
- Quality Data: Ensure clean, useful, well-governed data—“garbage in, garbage out.”
- Right-Sized Models: Consider small language models (SLMs) for domain-specific workloads; easier to train, tune, and deploy.
- Infrastructure: Start small (laptops, desktops, workstations with GPUs like Dell + NVIDIA), but ensure seamless scalability to larger compute when needed.
- Combination of Hardware/Software: Use the right stack for your workflow; Dell + NVIDIA offer free tools and development hardware (e.g., GV10).
- Governance (COE): Centralized strategy is critical for sustainable, organization-wide AI transformation.
Notable Quotes & Memorable Moments
- “A simulation that would have taken hours or days is 10x, 20x, or even 50–100x faster, depending on the application.” — Himanshu Iyer [07:57]
- “Agentic AIs are assisting designers and engineers at the stage they are working on for the manufacturing process.” — Himanshu Iyer [14:36]
- “To benefit from AI ... there needs to be that digital thread connecting all these different models and approaches.” — Himanshu Iyer [23:20]
- “Where we are right now, these LLMs don’t necessarily fully understand the laws of physics ... new models will be physically aware.” — Himanshu Iyer [26:40]
- “Doing all of this testing or simulations physically is impossible...doing all of this virtually in physics AI models is the only way we can move forward.” — Himanshu Iyer [34:38]
- “It starts with data—make sure you have good quality data ... garbage in, garbage out.” — Himanshu Iyer [36:54]
Key Timestamps
- 02:59 — Overview: Complexity of manufacturing processes
- 05:57 — Simulation shift: From CPU to GPU-enabled workflows
- 09:33 — Democratization of simulation: Workstation access
- 12:29 — Data, silos, and consolidation via AI/digital threads
- 14:36 — Agentic AI and iterative experimentation
- 17:02 — Scaling AI from PoC to wide deployment, organizational structure
- 21:39 — Center of Excellence vs. decentralized AI adoption
- 25:47 — Introduction and definition of physical AI
- 29:41 — Imparting physics into AI models; accuracy in synthetic training
- 33:41 — Benefits and necessity of physics AI/robotics in manufacturing
- 36:54 — Five key steps to prepare for AI transformation
Closing Resources
- Connect with Himanshu Iyer: LinkedIn [40:08]
- NVIDIA Developer Tools: build.nvidia.com [40:17]
- Dell Pro Max Resources: Linked in episode description
Tone and Takeaways
The tone of this episode is accessible, candid, and insightful. Logan plays the curious learner while Himanshu provides deep technical expertise, breaking down complex ideas into practical, actionable guidance. Manufacturing organizations at all stages—novice or advanced—will find the roadmap laid out practical: invest in data quality and AI governance, stay agile with infrastructure, and embrace evolving models from generative to physical AI for maximum workflow transformation.
