Ask HN: How do you train and deploy machine learning models at your company?

14 points·geoffxy··5 comments

Hey HN!

I’ve been reading a lot about ML and I’ve been curious: how do you train and serve models (for example, deep neural networks) at your company? Do you use a custom system or something else? How do you deploy/serve your models in production? Are there any particular pain points with your current approach?

Thanks!

View on HN

Curious as well. Have been looking at stuff like https://www.floydhub.com/ and https://www.datarobot.com/product/ - we have some use cases we want to implement but nothing built yet.

Flask/Docker/Postgres.. bit of a learning curve but more than enough resources online to get going.

so you save your model weights to Postgres, expose the model via an HTTP api using flask and package it w/ docker?

Full disclosure: I am a data scientist not a production-grade SWE.. in any case that sounds about right. I would probably just leave the weights loaded in memory which might be lower latency than querying a db every time you want the score. The DB is mainly there to store the json requests (or however you're receiving them) as well as the predictions for future use.

Another use case for the db would be if someone sends you data to score and you have to append other external data tables to that data, in which case you could use the db to append the data.

not really a company, but for my (profitable) side projects I just serve the models with flask/wgsi, on a single server.

its much cheaper than the cloud solutions, but I suspect it will fall over if it ever gets on hn..

Reply on news.ycombinator.com