A Former Nvidia Engineer’s Guide to Improving AI Skills

A former Nvidia engineer says those looking to stay competitive in the AI ​​era need to start building now.

“It doesn’t matter what you build, as long as you build it end-to-end: starting with an idea and deploying it for a friend to use,” Chip Huyen, an AI researcher who worked on Nvidia’s NeMo platform and taught machine learning at Stanford, wrote to Business Insider.

Huyen’s advice extends beyond his fellow AI researchers and engineers. She said that when she does workshops for businesses, it’s amazing to see what people without even a coding background can do, thanks to AI coding agents.

“After that, they have a lot more confidence and also have a much better understanding of AI,” she said.

If you’re not sure what to do, Huyen says she has a simple exercise.

“For a week, write down everything that frustrates you,” she wrote in a guide she shared with Business Insider. “Why is this taking so long? Why do I have to repeat this every time? etc. For each of these problems, try to think about how to solve it. Then choose a problem to solve.”

It’s also important to know the fundamentals of AI and how it works, Huyen said.

“Learning only by building is like learning a new language only by speaking,” she wrote.

That’s why it’s imperative that those looking to hone their skills complement their own exploration of AI by educating themselves about the tools they use.

“You also need to structure your learning, for example by choosing a program, books, courses,” she wrote.

If you’re looking for where to start, Huyen shared a list of AI-related resources with Business Insider that she compiled while writing her latest book, “AI Engineering.”

What Computer Science Students Should Do

As for those studying computer science or engineers looking to expand their knowledge base, Huyen said the best thing they can do is master systems thinking.

“Coding is just a means to an end. CS is about systems thinking, using coding to solve real-world problems,” Huyen said on a recent episode of “Lenny’s Podcast.”

Problem solving will never go away, Huyen said, because as AI can automate more things, the problems will only get worse. »

Huyen said AI “can help automate many disparate skills,” but it will remain difficult to combine distinct knowledge to solve problems. She showed a webinar hosted by Mehran Sahami, chair of Stanford’s Department of Computer Science, and Andrew Ng, founder of DeepLearning.AI and former founding member of Google Brain, during which the two men discussed the future of CS. At the time, Sahami said it was “critical that students understand the code generated” by AI tools.

In talking to companies, Huyen said she hears examples of senior engineers spending more time reviewing code written by junior engineers. Across tech, the fear is that more junior engineering roles will be replaced as generative AI becomes more proficient at writing code itself. That’s why skills development can be helpful, Huyen said.

Companies “value engineers who have a good understanding of the whole systems and are capable of good problem-solving skills, who think holistically rather than locally,” she said.