Mystic AI (YC W21)
I joined Mystic AI (YC W21) as employee #3, back when the company was called Neuro AI
. Like a precursor to FAL or Modal, we were building an API for generic ML compute
. I was a wildcard engineer: deploying models
, developing websites, managing servers (we had an on-prem GPU farm), writing docs, sorting through candidates, designing adverts
, and running demos. 1,200+ commits across 30+ repositories in 19 months. I single-handedly wrote and deployed every public inference endpoint
on our API, becoming one of the first in the market to help companies use open-source models
like GPT-J
in production. I worked on the Model Hub and the Pipeline dashboard from the first commit. As my own marketing play, I created playgrounds.ai
, a free inference UI for generative models like DALL·E Mini. From GPU allocation in Python to Kubernetes infra handling, from figuring out CUDA kernels and quantisation to implementing DeepSpeed, I worked on everything, while being the main point of contact for all users too. Additionally, I wrote our docs, newsletters, and emails (and I built custom platforms for managing them all). I started partnerships between Mystic and EleutherAI, The Eye, PhilPapers, and Georgia Tech.
