Keith Stevens

With 20 years practicing Buddhism and 15 years working in Natural Language Processing, I I’m convinced that language technology works best when it helps people live economically better lives so they can spend more time building and enjoying the human relationships that give life meaning. More so, the best solutions come from teams that focus deeply on a poignant market problem and solve it with relentless focus. And that’s what I hope to keep doing.

Over 9 years I worked with the Google Translate team while we did exactly that for machine translation. We got to live through the paradigm shift from old school statistical NLP to large transformer models, being one of the first teams in Google to productionize transformer models. With my direct teams we built systems that discovered, curated, and housed the vast pools of data needed to train these models. And with a small team I personally led we build out user facing features that let users share data and insights with us.

I then left Google and spent about 2 and a half years seeing the startup world up close. I tried replicating some of my Google work as an open source project. Partnered with a co-founder to try making an emotional intelligence oriented AI startup. I’ve since been working with several startups to build out their LLM infused products. Until mid 2024, I plan to continue doing this will splitting my time between Tokyo and Hakuba Japan. In my spare time I’ll also be skiing and writing mini reports on various LLM adjacent open source projects that catch my interest.

Now tho, I’m back at Google working on Gemini in the Bay Area. That’s included some fun work like ensuring Gemini’s data science like capabilities make life easier for anyone that hates using Excel and work that is soon to released. All this while my wife is doing her MBA at UC Berkeley.