Deep Learning classifier services via AWS Chalice. It's trivial. However, 50MB/250MB lambda limit is a pain, one has to cut down TensorFlow, Keras etc. signific

Deep Learning classifier services via AWS Chalice. It's trivial. However, 50MB/250MB lambda limit is a pain, one has to cut down TensorFlow, Keras etc. significantly before deployment (doable but difficult) or do some S3 tricks with /tmp. I wish they allowed increasing this limit for extra money. It's cheaper than EC2 Elastic Bean Stalk though. I wouldn't do that again; I can waste time on more interesting things than to hack artificial limits of architecture that will be changed at some point anyway.