Postdoc: Watermarking AI Audio & Speech Tech

Aalto University; Espoo; Hybrid; EUR 40,000 - 54,000; Full time

Benefits offered by this job

Hybrid model

Open science and open-source releases

Excellent computing infrastructure

Job summary

Aalto University in Espoo invites applications for a Postdoctoral Researcher in Speech and Language Technology. The role focuses on developing deep content-based watermarks for audio and collaborating on robustness, evaluation, and generalization with a doctoral researcher.

You will work on watermarking methods for generated speech and audio, design perceptual quality metrics, and publish in top venues, with access to excellent computing infrastructure and an international network.

Qualifications

• A doctoral degree, or close to completing a doctoral degree, in speech or audio processing, machine learning, signal processing, computer science, or a related field.
• Strong experience with deep learning and generative models.
• Programming skills suitable for modern machine learning research, preferably including Python and deep learning frameworks.
• Interest in one or more areas: speech synthesis, music generation, neural audio codecs, diffusion models, audio language models, watermarking, deepfake detection, audio quality assessment, or audio representation learning.
• Ability to conduct independent research and publish in high-quality peer-reviewed venues.

Responsibilities

• Develop new content-based audio watermarking methods for generated speech, music, and audio.
• Design detection and perceptual quality measures for perceptual integrity when generated audio deviates from reference content.
• Explore methods such as audio content detection, music information retrieval, audio fingerprinting, alignment search, and self-supervised representations.
• Integrate watermarking capabilities into deep audio generative models, including diffusion models and audio language models.
• Study embedding watermark information in long-term content like semantics, duration, prosody, or emphasis.
• Develop robustness evaluation methods, differentiable augmentation tools, and realistic attack scenarios.
• Publish research in leading journals and conferences and contribute to open-source releases.

Skills

Deep learning

Generative models

Python programming

English fluency

Research collaboration

Education

Doctoral degree in speech or related field

Tools

PyTorch

TensorFlow

Git

Job description

Aalto University in Espoo invites applications for a Postdoctoral Researcher in Speech and Language Technology. The role focuses on developing deep content-based watermarks for audio and collaborating on robustness, evaluation, and generalization with a doctoral researcher.

You will work on watermarking methods for generated speech and audio, design perceptual quality metrics, and publish in top venues, with access to excellent computing infrastructure and an international network.