Coffee Chat #25: What separates good engineers from great ones?

‍Q: You’re based in Silicon Valley, one of the world’s most intense tech hubs. What’s it actually like working as an engineer there, and what makes it different from other places?

A: The biggest difference is the concentration of talent. The Bay Area attracts talent from across the globe, often people who have fought hard just to get here. That means you’re not only surrounded by skilled engineers, but by people who are deeply driven, hardworking, and hungry to keep going. And that concentration of talent is self-reinforcing: talent goes where most of the jobs are concentrated, and the more talent there is, the likelihood that more companies follow. It’s a cycle that feeds itself.

That density cuts both ways, employers have unmatched access to skilled talent, and engineers have far more career options. Although how many depend on the state of the market.

But here’s the thing I find most interesting: engineering work, even work on supposedly “disruptive” technology, is almost always gradual. The person doing it is usually just picking up where the last person left off and pushing things a little further. It doesn’t feel disruptive when you’re the one doing it day to day. What a tech hub like the Bay Area really enables is that those small, incremental steps happen much faster. The density means adjacent problems are being solved all around you, and that naturally accelerates everyone’s thinking. Many small steps compound into something that looks disruptive from the outside, but the real engine underneath is just a higher rate of gradual progress.

Q: Having worked with teams across the globe, what is one habit or skill that distinguishes a “great” engineer from a “good” one, regardless of where they are in the world?

A: Curiosity about the actual problem, not just the technical solution. I’ve seen plenty of engineers who get handed a task and immediately start coding, and end up building something technically impressive, but that does not address the needs of the customer.

A great engineer asks “why” before “how.” Why does this defect matter? Why is the customer measuring it this way? Why does this particular scratch on a part cause a rejection while another one doesn’t? Once you understand the real-world context, the technical approach almost reveals itself. Or at the very least, you avoid spending significant time solving the wrong problem.

This applies regardless of where you are located in the world. The language and culture might be different, but the engineers who take the time to understand the problem deeply, who go visit the factory floor, who ask the uncomfortable questions, are the ones who build things that actually get used.

Q: One of the biggest hurdles in AI is having enough data. Your company works with tools that create “synthetic” defects to train AI. How is this changing the speed at which engineers can deploy solutions in the real world? ‍

A: In our industry, we’re actually incentivized to train on as few images as possible. We can gather plenty of data in our own lab and build strong base models, but when we deploy an inspection system at a customer’s site, we can’t sit there collecting thousands of images. Usually, a customer only has a handful of examples of a particular defect, often in the single digits. And if that defect only appears in one out of every ten thousand parts, you’d need to run fifty thousand parts just to collect five examples. That’s not realistic when someone wants a solution to be deployed quickly.

This is where synthetic data generation has been massively beneficial for us. A customer can show us one, two, three examples of a defect, and we can use generative models to create realistic variations, different sizes, orientations, severities, and lighting conditions. We then verify those generated samples together with the customer and use them to train the smaller, specialized networks that run in production. These aren’t massive general-purpose models. They’re fast, purpose-built networks designed to meet strict cycle times on the production line.

‍Q: You lead Build Projects at Open Avenues where students transform 2D images into 3D data. Why is “3D thinking” a skill that will be essential for the engineers of the 2030s?

A: 3D feels intuitive to us as humans because that’s how we experience the world. You can look at a scene and tell that a door is behind a person or roughly estimate the distance to a chair. Most industrial robots today are what we’d call “stupid”; they do repetitive tasks extremely well, but they are blind.

There’s a growing set of problems that can’t be solved without vision, and in many cases, 2D vision isn’t enough. You need to understand depth. How far down should a robot move to pick up a box? How deep is that scratch on a surface? These are fundamentally 3D questions.

What I teach in my project is essentially how a traditional 2D camera actually sees in three dimensions. How that projection works, how a camera fits into a 3D coordinate system, and how you can work with cameras inside larger systems like robotics. We use stereo vision to generate 3D as a core concept, but the principles apply broadly to many methods.

‍Q: What has surprised you most about working directly with customers in manufacturing, and what can students learn from that?

A: The thing that surprises me every time is the sheer variety of problems that exist in manufacturing that I never would have imagined. You think you know what “defect inspection” means, and then you visit a customer’s facility and discover they’re dealing with something you’ve never seen before, or that they want to use our technology for something we didn’t even realize was a problem.

That’s what excites me about this work. I get to collaborate with people who have deep expertise in their own domain, battery manufacturing, consumer electronics, automotive parts, and they come up with real problems that need innovative solutions. Together we figure out how to solve their specific challenges using our Vision and AI technology. I never deal with the same problem twice, and I get to see the real-world benefit of the solutions we build.

For students, the lesson ties back to what makes a great engineer: listening to your customers. Understand what they’re actually struggling with, understand what they want to achieve, and then help them solve that problem in the best and most efficient way. The best engineers I work with aren’t the ones who know the most algorithms. They’re the ones who can sit with a factory manager, understand what keeps them up at night, and then go back to the lab and build something that actually helps. If you can develop that skill early, you’ll stand out in any field.