Duke Lee

Software Engineer · Infrastructure & Distributed Systems

Professional Experience

Engineering Stories

How I helped uncover malicious ads that were hiding from our scanners

At Meta, some malicious ads hid harmful links behind images or buttons that our scanners didn’t interact with. I spent six months building a crawler that could follow those interactions and uncover links the existing system missed. I started with ads that had already been reported, then learned another team had found a way to identify suspicious ads earlier. Bringing their work into the project showed me how much the scope of an engineering problem can change when you talk to the right people early. The crawler became part of Meta’s advertising security infrastructure, which processed millions of ads a day, and uncovered malicious content that was later reported and blocked.

What building for reliability taught me about developer experience

At Twitch, I led a small team developing reliability infrastructure to help services stay operational during unexpected traffic spikes. A single overloaded service can cause failures elsewhere, so we gave teams shared tools to limit incoming requests and shed excess load. The tooling worked, but adoption was lower than we’d hoped because setup was cumbersome and it took effort to see the benefits. That experience taught me to design for the first few minutes of using a tool, with sensible defaults and approachable documentation.

A Little About Me

Outside of engineering, I enjoy exploring my creative side. I started painting at 25, with no formal training, simply because I wanted to try something new. Since then, painting has become a way for me to express myself in ways that words sometimes can't. I particularly enjoy creating portraits, including a self-portrait. I also enjoy video games, particularly competitive fighting games like Tekken, and occasionally experimenting with small creative projects of my own.

Other Things I’m Working On

Afterglow

Afterglow is a movie-search project I’m building that lets you describe what you want to watch in ordinary language—for example, “a funny movie from the 90s with dogs in it.” It uses an LLM to interpret the request and aims to find relevant movies.