As stated above, querying with case-insensitivity and truncation is very slow in SPARQL query service.
I found this project on github: https://github.com/inventaire/entities-search-engine
It sets up an ElasticSearch index which allows fast queries for use-cases like autocompletion.