I have developed many machine learning and deep learning models in the past year for my startup. We found many resources to easily train and make good models.
However, when we needed to deploy those models to production and use it on a website/mobile app, there were very few resources. We had to wrap our models as REST API using Flask, serve it using gunicorn, download all the dependencies, containerize it using docker and deploy it on kubernetes. We also had to use NGINX to handle multiple requests.
This entire process was very time consuming and for someone like me with very little backend development knowledge, it was a big pain to set up. Even after we set it up, it wasn’t easy to maintain it and we encountered many scaling issues.
After my last startup failed, I and my co-founder decided to work on developing a platform to make it easy for data scientists like me to deploy models to scale. We are thinking of developing a platform where data scientists can just upload their trained models and we provide them with an API for inference. All the scaling and management of requests will be handled by us.
However, we are not sure if deployment is a big problem for all data scientists and AI startups or it’s just us. If you are someone who trains models and finds it difficult to deploy them, can you please sign up as a beta-tester for our platform: https://goo.gl/forms/epTKgUStG6MEx6xi2
If we are able to sign up enough beta-testers, we will definitely build the platform. Share it with anyone else you know who this might be of help as well.
Thanks,
Prasann