ASR / speech recognition expert for a custom ASR service

I need an expert in ASR / speech recognition to create a custom ASR service for us based on Mozilla DeepSpeech (or similiar open source technology - this is up to you). The language recognition should be in GERMAN. ONLY GERMAN! You should use all the online available training models to get the best results. Additionaly I can offer a database with 25 hours of high quality speech files + text for training as well. Please only apply if you have already built and operated such systems yourself. In the end you should have an application that runs on my own servers, does not depend on external services and provides reliable and fast speech recognition. But there is one special feature: For each audio file sent, the expected text is available. So there are two input parameters: The audio file and a string with the expected content. The program only has to reliably determine whether the words in the string also occur in the audio file (in the correct order, of course, preferably with a score). The best result would be the following: word1 : 1.0 (best recognition result) word2 : 1.0 (best recognition result) word3 : 0.5 (not sure result) word4: 0 (missing) So it is not a direct speech recognition, but rather a check, because the expected result is always known. I offer a VM (ubuntu) in a data center to run the software. I would prefer a REST-API (http-post) for the requests.