Independent research
An independent, self-funded comparison of speech-to-text systems, local models and cloud services alike, run on identical audio and scored with one published method. It looks at how accuracy and behaviour change with the audio: clean speech, background noise, telephone lines, parliamentary debate and longer recordings, across several European languages.
No vendor sponsored it, reviewed it or saw it before publication.
I am a senior solutions engineer based in Zurich. For more than ten years I have worked in technical pre-sales, helping organisations find out whether a technology fits their problem before they commit to it: discovery, technical validation, workshops and proofs of value.
I currently work in conversational AI, with a focus on voice applications that have to work across languages. Italian is my first language, and I work in English, German, French and Spanish, which is why my speech-recognition testing is not limited to English.
Alongside work I am studying for the MAS ETH in AI and Digital Technology at ETH Zurich.
A personal activity, separate from my professional work and from the benchmark.
The best way to reach me is a message on LinkedIn.