SF Big Analytics - System Design for RecSys & Search

Eugene Yan, who writes some of the more practically-minded machine learning content out there, took the stage at the San Francisco Big Analytics meetup this July to talk about something that rarely gets a clean explanation: how you actually architect a recommendation or search system end to end.

Most people learn RecSys and search from scattered blog posts or research papers that focus on one narrow piece, a ranking model here, an embedding trick there. Yan's talk instead walks through the full pipeline as a system, the kind of thing you'd sketch on a whiteboard before writing a line of code. That's a different skill than tuning a model, and it's one a lot of engineers pick up the hard way, by shipping something that falls over in production.

The talk is now recorded and public, alongside the slide deck, so anyone who missed the meetup can watch it on their own time. There's no paywall or account wall mentioned, just a straightforward link to the recording and the slides for people who want to follow along or reuse the material.

Yan even included a formal citation format for the talk, in both a plain reference style and a BibTeX entry, which is a small but telling detail. It signals he expects this to be referenced in other people's write-ups and papers, not just watched once and forgotten. For a field where system design know-how is usually passed down informally, having a citable, recorded source like this is a minor but useful contribution.