This week in Silicon Valley, I had the chance to see some of the most influential companies up close and get a real sense of how technology is evolving. The trip gave me a clearer picture of where a number of industries are heading, from breakthroughs in AI and robotics to conversations with people working on these advancements. Let's break it down.
Automation
Tesla
The Tesla factory in Fremont was a huge reminder of the fact that assembly and manufacturing technology is always changing and increasing in precision. One thing that especially stood out to me about the Tesla factory was the usage of space. On one hand, there were extremely constricted, fully automated pipelines for Model 3 assembly, where robots were on the ground and hanging upside down in a mere 6 foot tall box manufacturing the car frame end to end. On the other hand, there were machines that lifted car chassis up 6 feet, just to move it across horizontally, before lowering it down 6 feet again, to save floor space in between. Now, I'm not going to presume to know more about robotics and manufacturing than the experts working at Tesla, but I will say this. The ground footprint and the space we take up is so extremely valuable. I would go so far as to say we might even see vertical assembly lines one day, to save on square footage.
The other thing that struck me was how different the assembly lines were from model to model. It was very different from what I noticed in Bloom Energy (see below), in the sense that there didn't seem to be a uniform integration or methodology in using the very best procedure available to the plant. Perhaps the restructuring would be too costly, but that is also a very well known fact about Tesla's company culture, in that they make changes quickly and constantly.
Bloom Energy
Now, Bloom Energy was another company we visited, and the key message that was emphasized over and over was modularity. Their product revolves around providing modular, secure, disaster-resistance hydrogen power to clients. This same ethic, similar to Tesla's of constant change, was seen in their factory and innovation procedure. We were outlined a very systematic procedure about how new manufacturing technologies were integrated into their pipeline.
- A new fab would start in Stage A, then after testing for accuracy, would move its way to Stage B.
- After being tested for robustness / effectiveness for mass production, it would move into Stage C, where it would slowly be integrated into the main product line.
- Finally, after some practical usage, it would be moved to Stage D, where the entire old product line would be replaced by the new (and presumably better) technology.
The future of automation
Honestly, I found the automation style of the two companies very telling in regards to their company values. At Tesla it's all about change, constantly modifying the way they work as well as the end product. On the other hand, at Bloom Energy, it was all about minimizing defects, and precision manufacturing can't afford to make big, fast changes.
One thing I believe will be a major player in the future of automation is simulation of new machines. Especially with tools like NVIDIA Omniverse coming out, it remains a frontier that, if properly tested, could revolutionize the process by which companies go from an idea to a large scale implementation.
Quality Control
There's a huge movement in AI startups to doing more with logic, agents, and overall decision-making and problem solving. However, many established companies are still trying to incorporate AI technologies that many newer startups overlook into their existing lineups.
ASML
One of ASML's biggest drivers of revenue is repairing defects in their lithography-produced microchips. A major part of the San Jose office is, in fact, dedicated to this task. With that being said, there's a lot of manual labor by experts that goes into identifying defects and trying to determine causes. That's why ASML is having a major push toward incorporating AI into defect detection and reasoning.
It's not like we haven't seen this before; I mean, medical AI is huge. AI tech promising to identify defects in the human body, or at least assisting doctors in doing so, is a task that many medical tech startups have set their eyes on. Yet, the human body is so complex and so diverse. Imagine, instead, that you look for defects in a regularly produced, data-driven, precise microchip. Of course this is a far more promising application of defect detection! I just feel as though it's not talked about enough. With advances in defect detection, companies like ASML have a golden opportunity to intensely speed up their iteration process on fixing lithography outputs and boosting revenue.
AR/XR
This is probably the first blog post I put up, but if you keep reading, hopefully I talk about my belief that Augmented Reality (AR) and Extended Reality (XR) are going to be a huge player in the everyday tech space. I think that they will be synonymous with the smart phones we have today, even going so far as to replace many of their use cases. But honestly, most of the AR/XR hype seems... far off to say the least. At least, that's the impression you can get by reading tech news. Let's be honest, the most comprehensive AR product on the market is probably the Apple Vision Pro, simply because of its features. But it's bulky and slow.
In any case, the technology of easy to use, everyday wear, AR glasses seems closer and closer by the day. It's not so much the actual hardware that was interesting to me, it was the development platforms and operating systems. I know it's easy to write off the various OSs as a foregone conclusion (Apple will continue iOS, Android OS from Google, etc.), but it would be so amazing if there was a new operating system from a third party that integrated with AR better than the existing duopoly.
The Future of Robotics
The progress in robotics was another highlight, especially in terms of simulation and decision-making. At NVIDIA GTC, I saw how companies are training robots in virtual environments before deploying them in the real world. This allows for fast, large-scale testing without the costs and risks of real-world failures. Training in a simulated environment means robots can go through millions of trial-and-error cycles in hours rather than weeks, refining their behavior before they ever interact with physical objects.
Now, I talked about this before, but simulation is going to be the forefront of AI in the coming 5 years. I was blown away by a specific platform in particular: NVIDIA COSMOS. COSMOS is NVIDIA's simulation engine, using machine learning models that understand physics in the process. It's a truly innovative system, and I think as research iterates on simulation platforms like this, we will see leaps and bounds of autonomous reasoning being used in the real world.
My Thoughts on AI Education
One thing that stood out across all of these companies was how specialized AI has become. Right now, many AI courses treat deep learning as one big concept, but in reality, there are distinct areas like computer vision, language processing, robotics, and reinforcement learning, each requiring different approaches. The industry has moved far beyond general AI models, and education needs to reflect that shift. I think that the focus in AI at many universities that purportedly have stellar AI education is focused into a specialization within a larger Computer Science major, instead of treated seperately. This makes students have their AI education restricted to just 1 or 2 years, meaning they have to pursue a graduate degree to make meaningful contributions to the field (not just using APIs or wrappers). As a student at Purdue University, I think that the AI major that we have here illustrates this difference quite well. AI students must take basic programming classes, as well as theoretical foundations of CS classes. But, instead of focusing on computer architecture or systems programming, they instead focus on algorithms used in AI, deep learning, and have much more room to explore new technologies in electives.
Universities that fail to adapt to the push for AI experts in industry will leave students underprepared for the roles companies actually need to fill. Even when looking at Purdue’s AI major, I can see that it lacks concentrations or specialized courses despite AI being a rapidly growing field. The fact that the deep learning class is a 500-level course makes it inaccessible to many undergraduates, limiting early exposure to these critical concepts. If a CS major can choose to specialize in cybersecurity, software engineering, or systems, an AI major should be able to specialize in computer vision, reinforcement learning, autonomous robotics, and so much more.
What's Important
At the end of the trip, I left with a better understanding of what makes these companies successful. It’s not just about inventing new technologies—it’s about making sure they work in the real world. The importance of integration, specialization, and continuous learning became clear across every company I visited. Seeing these ideas in action made me even more excited about the future of AI, robotics, and automation, and I’m looking forward to finding ways to contribute to these areas myself.