Comment by schmookeeg

I took a recurring chore at my last day job — optimizing indices against large embedding datasets — and made a little SaaS-y thing [1]. It auto-tunes my vectors and shows how/why the benchmarks shook out like they did, with prettier graphs than I was spinning in Jupyter.

I use it to learn about weird, difficult datasets to optimize for. I've been at it over a year, and have met some interesting people in this obscure niche. It keeps me off of the streets. ;)

[1] https://www.vectorpanda.com/