Kicked out of Columbia, this student doesn't plan to stop trolling big tech with AI

A computer science student at Columbia University said he has been kicked out by the school after he built an artificial intelligence tool to cheat in tech job interviews and documented the fallout online.

Chungin “Roy” Lee, a second-year undergraduate, garnered online attention after he claimed to have fooled four of the world’s biggest companies using Interview Coder, a desktop app he created to discreetly solve technical coding questions.

In a now-removed YouTube video, Lee 21, recorded himself using the tool during an internship interview with Amazon. His app, which he said took only four days to build, allows users to take screenshots of problems without being detected by their browsers. It then processes the images using AI to spit out solutions in real time.

It was a stunt that highlighted the proliferation of generative AI technology in everything from schoolwork to technical jobs, as users discover new tools to help them cover for their lack of skill or knowledge or to otherwise enhance their abilities. As such tools advance, schools and workplaces have struggled to accurately detect their use.“I think 99% people probably haven’t realized how far-reaching this could be,” Lee told NBC News. “In the past, you could have built an invisible desktop assistant, and you also could have used LLMs [large language models] to solve problems. But now that people are putting the two together, I think no form of online assessment is safe.”

Lee, who is in New York City, said he got the internship offer from Amazon this year. In February, he went online to broadcast that he “used AI to pass my Amazon Interview.” (The video got about 100,000 views before YouTube removed it, citing a copyright claim by Amazon.)

Lee claims he used Interview Coder to nab offers from all four of the companies he tested it on: Amazon, Capital One, Meta and TikTok. He said that he rejected Amazon’s offer and that several of the other companies rescinded their offers after he publicly posted about his use of AI.

As a result of his publicized stunt, Columbia University ushered Lee through a disciplinary process that resulted in a yearlong suspension, he said. A spokesperson for the university declined to comment on individual students, citing Family Educational Rights and Privacy Act regulations.

Margaret Callahan, a spokesperson for Amazon, said that while the company welcomes candidates’ sharing their experiences working with generative AI tools when it’s relevant to the roles they’re applying for, they must also acknowledge that they won’t use unauthorized tools — such as generative AI technology — to help them during the interview or assessment process.

Capital One, Meta and TikTok didn’t respond to requests for comment.

Since he began documenting his journey on social media roughly a month ago, Lee said, Interview Coder has had user growth of 50% week over week, and it’s now making $170,000 in monthly subscriptions.

He has continued building his brand online as a combative rule breaker by posting what he claims to be emails with recruiters and documents from his subsequent disciplinary process with Columbia. He has gained more than 43,000 new followers on X, where he has seen the most support for his stunt. On LinkedIn, where he has also gained about new 11,000 followers, the reactions have been more mixed, with some users calling his product “morally unethical” or “corrupt.”

Lee said he’s “not really worried about burning bridges” within the academic and corporate worlds at the start of his career, as he prefers to invest in entrepreneurship.

Trolling the companies, he said, was an act of protest against what he perceives as their overreliance on LeetCode, a popular online platform that provides preparatory questions for coding interviews.

“I was in the top 2% of global LeetCode users, and this was probably, like, the 600 most miserable hours of my life when programming,” Lee said. “The questions are not indicative of what you do in the real world. They’re riddles that you have to memorize.”

Other AI-powered interview cheating apps, such as Leetcode Wizard, which touts successful results for thousands of users, have also become increasingly popular during the generative AI boom of recent years.

Some of them specifically tap into what many software engineers have expressed as a growing disillusionment with LeetCode.

LeetCode “doesn’t properly allow engineers to demonstrate their abilities when it comes to programming, while also ignoring the use of AI in daily programming work,” Isabel de Vries, Leetcode Wizard’s head of marketing, said in a statement. “It’s understandable that candidates would want to use similar tools during their interviews.”

Lee said he knew posting his Amazon interview would be a “really visual, viral way” to tackle the issue that motivated him to create Interview Coder in the first place. He said success would mean getting big tech companies to change their interview formats. Now that tools like his are rising in use and popularity, the genie can’t be put back in the bottle.

“I think cheating on these online assessment interviews is really much more common than people think,” Lee said. “Many CS students who are pretty serious about recruiting and stuff know that there are apps like this. It’s just not understood at a large scale.”

This story first appeared on NBCNews.com. More from NBC News:

• HHS plans to shutter or downsize several health agencies, including at CDC
• Former national security officials say plans shared with journalist should have been classified
• Utah governor signs bill to phase out state's universal mail ballot system