Hello world! đź‘‹
I’m Michele Mancusi, a Senior Applied Scientist at Music.AI (Moises), where I develop and optimize autoencoders and generative models for speech, music, and general audio enhancement.
Previously, I was a Research Scientist at Sony, where I conducted research on deep-learning-based generative models for speech, audio, and music, including Large Language Model (LLM) and diffusion-based approaches.
Earlier in my career, I interned at Microsoft and Musixmatch. At Microsoft, I worked on deep learning for unsupervised speech separation, while at Musixmatch I focused on singing voice detection.
I earned my Ph.D. from Sapienza University of Rome under the supervision of Prof. Emanuele RodolĂ as a member of the Gladia research group. My doctoral research centered on music generation, source separation, and Natural Language Processing (NLP), contributing to advancements in the field of generative AI.
- Deep Learning
- Signal Processing
- Generative AI
- Music Generation
- Source Separation
- NLP
- Speech Synthesis
-
PhD in Computer Science, 2024
Sapienza University of Rome
-
M.S. in Physics, 2019
Sapienza University of Rome
-
B.S. in Physics, 2016
Sapienza University of Rome
Work Experience
- Developing and optimizing autoencoders and generative models for speech, music, and general audio enhancement.
- Conducted research on deep-learning-based generative models for speech, audio, and music, including LLM- and diffusion-based approaches.
- Worked with Dr. Stefan Uhlich in the AI, Speech and Sound Group. Conducted research on deep learning for effects removal and timbre transfer with diffusion models.
- Worked with Dr. Sebastian Braun in the Audio and Acoustics Research Group. Conducted research on deep learning for unsupervised speech separation.
- Worked with Dr. Loreto Parisi in the AI Team. Conducted research on deep learning for singing voice detection.
Publications
Achievements
- First-author paper selected among the top 1% submissions for an oral presentation at ICLR 2024
- Awarded €50,000 in AWS credit for the best generative AI research project
- Awarded €20,000 for the best machine translation research project
- Recognized as one of the top doctoral research projects and awarded research funding
Academic Experience
- International Symposium on Diffusion Models for Audio and Music Processing
- OsnabrĂĽck, Germany
- Lectured at the Deep Learning and Applied AI MSc course and mentored students for their Master’s Thesis
- Delivered a guest lecture on the Latent Autoregressive Source Separation paper upon the invitation of Prof. Ronald Coifman
Attended
Skills
