PunySquid88 I'm not very familiar with what they do, but it seems that although it has a backend server as an option, it will mostly target single users with what seems like an easy to use single app.
From the Reddit thread it seems that it is still not open source, although they say they will...

Also from checking the examples it seems that like most experiment managers the integration is manual, you have to manually report a the stats, where as Trains does everything automatically. And lastly there is no DevOps solution, i.e. the trains-agent that allows you create your own cluster send experiments scale to the cloud and all without a single yaml file :)

A few more things that come to mind:
Unified Storage access , with built-in cache
Automatically create model file copy on a shared storage
Artifact management,
Query engine (retrieve experiments results, artifacts and models from code)
Automation and pipeline (trains-agent)
Hyper parameter search (soon to also include hyper band)
Resource management
Cluster creation with Docker support
Integration with all major cloud providers
K8s integration
Probably a few more things I forgot :)