Patrick von Platen

Created and led development of the Diffusers library -- the go-to open-source toolbox for state-of-the-art pretrained diffusion models for image, video, audio, and 3D generation in PyTorch. Over 33,000 GitHub stars.

Large-scale self-supervised model trained on 128 languages using 500K hours of speech data, achieving state-of-the-art results on speech translation, recognition, and language identification tasks.

Contributed the Wav2Vec2 model to Hugging Face Transformers and authored tutorials that made self-supervised speech recognition accessible to the broader community.

Co-authored at Mistral AI -- a 12-billion-parameter multimodal language model trained to understand both natural images and documents, with a novel vision encoder.

Robust knowledge distillation approach for speech recognition via large-scale pseudo labelling, enabling faster and smaller Whisper models.

Universal Stable Diffusion acceleration module that dramatically reduces inference steps while maintaining quality, enabling real-time image generation.

Evaluation benchmark covering 102 languages across speech recognition, classification, translation, and retrieval tasks for cross-lingual speech representations.

Co-authored the Datasets community library for NLP, providing 650+ datasets and tools for efficient data loading and processing, published at EMNLP 2021.