Modelbit

👋 Hey makers! My best friend Tom and I have spent our careers making data tools together. After spending the last decade building Periscope Data, we're launching our next project, Modelbit! 🚀 Modelbit is a labor of love after a decade spent watching data scientists struggle to deploy models. The data scientists at Big Tech companies get whole teams of ML Engineers to help them deploy. Everyone else is out of luck ... until now! From any Python notebook, just run `modelbit.deploy()`. From there, a whole bunch of cool stuff happens under the hood: 👉 Detect and containerize the running Python environment 👉 Ship the model and its container to the cloud 👉 Give the model a REST endpoint and Snowflake SQL function for inference 👉 Version control the model in your git repo 👉 Scaffold load balancing, logging, unit testing and more 👉 Stand up retraining pipelines, A/B testing, and feature stores in case you want them! We love data teams and we're proud of the work we do supporting and enabling them. We hope you'll find this latest project useful. ❤️

I already have a bit of context about Modelbit (we invested at Weekend Fund) but I'm curious to ask a questions somewhat unrelated to this launch... @harry_glaser – You and Tom previously built and exited Periscope Data for over $100M. You know how hard it is to build a large company. Why do it again? Isn't it easier to just rest and vest at a large co. 😅

After building many ML centric products over the years, I can attest that getting a model into production is one of the hardest problems in software. It's so hard that it often slows down product development to a crawl, as the Data Scientists don't have the skills to move a model into production but the engineers don't understand how the models work. ModelBit is addressing this head on and I wish we had it when I was still building products!