Research Scientist – Generative Audio

This is a fully remote position, open to applicants in New York.

📋 Description

• Engage in pioneering research within the realm of generative audio, utilizing diffusion or flow matching models, focusing on one or several of the following domains:

• Vocal Synthesis — Investigate vocal and speech synthesis along with related fields such as ML-driven audio processing and signal processing.

• Post-Training — Explore post-training methodologies for music generation, including preference alignment techniques (like DPO, RLHF, or KTO), the design and training of reward models, and reinforcement learning to enhance output quality, controllability, and adherence to human preferences.

• Editing — Conduct research in iterative music generation and audio editing, encompassing features like stem replacement, instrumentation alterations, mood adjustments (while retaining content), tempo modifications, and structural changes.

• Execute large-scale experiments utilizing Spotify's vast infrastructure and reach an audience of over 700 million monthly active users.

• Develop practical applications that leverage generative technologies and expand the possibilities of listening experiences.

• Collaborate within a cross-functional team—working closely with scientists, engineers, product managers, designers, user researchers, and analysts—to devise innovative solutions for complex challenges.

• Affect Spotify's products, tools, and services directly, engaging in projects that have an impact across the entire organization.

• Connect with the wider research community by publishing your discoveries, delivering presentations, and participating in leading conferences.

⛳️ Requirements

• You hold a Ph.D. in Computer Science, Mathematics, Engineering, or a related discipline. Previous industry experience is advantageous.

• You possess experience in one or more of the following areas: generative modeling, machine learning, music information retrieval, speech processing, audio processing, signal processing, probabilistic modeling, computer vision, or similar fields.

• You have profound expertise in at least one of the aforementioned focus areas—be it vocal/speech synthesis, post-training alignment techniques (such as PPO, GRPO, DPO), or audio-to-audio generation and text-guided music editing.

• You have published in esteemed conferences including ICASSP, ISMIR, INTERSPEECH, ICLR, AAAI, IJCAI, NeurIPS, ICML, CVPR, ECCV, ICCV, or similar venues.

• You demonstrate strong programming skills in Python, PyTorch, and NumPy.

• You are a creative problem solver who is dedicated to developing exceptional products that deliver real value to millions.

• You are excited about transforming research concepts into scalable products.

• You can simplify complex topics and enjoy fostering strong relationships with colleagues and stakeholders.

🏝️ Benefits

• Flexible work arrangements