PFISTERER
I grew up on a farm in Southern Germany.
At 17, I (cold) reached out to a billionaire investor and landed an internship in London.
Then, I moved to Zurich, to study computer science at ETH. My bachelor's thesis turned into EuroSpeech — a 61,000-hour multilingual speech dataset across 22 languages. It became a NeurIPS 2025 Spotlight and #1 trending on HuggingFace.
I'm currently based in Berlin and focusing on voice AI. Last fall I supported two voice AI investments at Balderton Capital. Now, I want to help shape the future of voice AI.
Voice is the most natural interface
I also like: (trail) running, (street) photography.
NeurIPS 2025 · Datasets & Benchmarks · Spotlight
S. Pfisterer, F. Grötschla, L.A. Lanzendörfer, F. Yan, R. Wattenhofer
A scalable pipeline for constructing multilingual speech datasets from parliamentary recordings across 22 European parliaments. Extracts over 61,000 hours of aligned speech in 22 languages, achieving an average 41.8% reduction in word error rates when fine-tuning existing ASR models.
Coming soon.
Coming soon.