A few years ago I told myself I’d start writing publicly “when I had something worth saying.” That turned out to be the wrong threshold — what writing actually does is generate the thing worth saying, by forcing the half-formed take in your head into something a stranger could read.
So this is me skipping the wait and starting.
A bit about me: I’m a Staff ML engineer at Meta, where I work on multimodal LLMs and large-scale recommendation systems. Before that, a PhD in CS at UW–Madison on causal inference and healthcare ML. Most of my career has been about making ML actually ship at scale and stay shipped — which turns out to involve a surprising amount of boring craft and a smaller amount of the interesting stuff that papers get written about.
What I’m planning to write here, roughly:
- Deep dives — long posts pulled from production experience. Scaling, ranking, multimodal, the gap between the paper and the system. The kind of thing I wish I’d been able to read before I had to learn it the hard way.
- Research notes — short writeups on papers and releases I’ve been reading. Less “review,” more “here’s what I noticed and what it changed for me.”
- Shorter takes — links, observations, quick reactions. The format Simon Willison has perfected. Keeps the site alive between the longer pieces, and is the part I’ll personally enjoy most.
No paid tier, no comments, no newsletter (yet). RSS is the canonical way to follow along.
If you’re reading this in the first week, hello — the cadence will be lumpy for a while. I’ll try to be honest about what I don’t know, and specific about what I do.