andrewgrowles

Read this the other day. I don’t see why Netflix needed to build nteract. JupyterLab has great extensibility (albeit very under-documented).

This is funny. My senior design project is very similar. We have a Steam VR app, built in Unity3D, where we can search/load 3D models at runtime from a backend web service. I've written distributed web crawlers (docker, headless chrome, selenium) to build a repository of 3D assets and used GC Storage for file storage. Then I have a web app with a search index (lucene based) that provides a RESTful api for the Unity3d client. It is surprisingly fun to create scenes in VR.

BigQuery would probably be the easiest, low maintenance approach. Normally I would leverage the Hadoop ecosystem for this type of ad-tech/analytics problem, but I am not 100% certain of your use case, budget, etc. Providing clients read-only SQL access would be much better than a REST API.

This is not going to perform well with this amount of data. HTTP Response times for queries are likely to timeout. Needs a job-based queue.

I'd look for a library and not write one from scratch. Lucene [1] and OpenFST [2] are great implementations. I haven't used C#, so I don't know if bindings exist or not.

Also you may find this talk useful [3] (Particularly slide 11).

Great write up by the way. Really thorough on the benchmarking!

[1] https://lucene.apache.org/core/4_1_0/core/org/apache/lucene/...

[2] http://www.openfst.org/twiki/bin/view/FST/WebHome

[3] https://www.slideshare.net/lucenerevolution/text-tagging-wit...

A FST seems like a good fit for this problem. I believe it will be much more compact than the Aho-Corasick algorithm trie structure. Depends on the size of the dictionary.