Amazon Science - Eugene Yan and the Art of Writing about Science
Eugene Yan
Amazon scientist Eugene Yan traded political trade agreements for recommendation algorithms — then became famous for how he writes about it. Turns out clear writing, not fancier models, is what actually gets noticed in ML circles.
Based on reporting by Eugene Yan — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Eugene Yan's resume reads like someone who kept changing his mind, except he didn't. He studied psychology in Singapore, then spent a stretch as a government policy analyst sifting through trade agreements and legal cases before deciding he missed numbers more than he expected. That itch took him to IBM in 2013, then to Lazada, the Southeast Asian e-commerce giant, then to a healthcare startup called UCARE.AI predicting chronic disease risk. Each job looked different on paper. Underneath, Yan says, the thread was the same: using data to help actual people, not just optimize a metric.
He picked up a master's in computer science from Georgia Tech in 2019 to shore up the fundamentals, then moved to Seattle in 2020 to join Amazon, where he now works on the recommendation systems behind personalized book suggestions on the Amazon Store. That's the job. The reputation, though, comes from somewhere else entirely — a personal site, eugeneyan.com, that he started almost as an afterthought.
He didn't set out to build an audience. Early mentors, several of them senior data scientists, kept telling him the same uncomfortable thing: technical skill matters less as you climb, communication matters more. Yan doubted it until he didn't. He started publishing on a barebones WordPress site, wrote dozens of posts nobody read, and then in 2020 relaunched as eugeneyan.com. One post on note-taking pulled 35,000 unique views in a day. Another, an opinion piece arguing data scientists should work more end-to-end, picked up 500-plus likes on Twitter. The snowball effect he describes is real: writing sharpened his own thinking, and the audience it attracted turned into a professional network he wouldn't have built any other way.
The chase for engagement burned him out fast, though. He tried writing for clicks for a while and found it hollow, so he pivoted to writing purely for himself — topics he wants to learn, pitched at an audience of people he'd actually want as colleagues. Fewer readers now, by his own admission, but the ones who stick around comment, argue, email him back. He says one real exchange beats ten thousand likes, and he means it literally, not as a humblebrag.
Inside Amazon, that instinct lines up neatly with the company's famously document-heavy culture — the working-backwards press releases, the six-page memos read silently before meetings even start. Yan says he writes almost as many documents as he writes code, and that the discipline of explaining a design clearly has become inseparable from doing the design well. His advice to anyone trying to build a similar habit is unglamorous: write for yourself first, and just start, because you won't find your voice sitting still.
My take — AI-written commentary, not fact-checked reporting
I cover AI news for a living and the thing nobody wants to admit is that most 'thought leadership' in this space is noise generated to farm engagement, which is exactly the trap Yan says he fell into and then climbed back out of. The lesson here isn't really about machine learning at all — it's that clarity is a moat, and in an industry drowning in hype-speak and recycled takes, the people who write plainly and for themselves end up mattering more than the people optimizing for the algorithm.
Read more about this at: Eugene Yan