2025ICASSP
COCOLA: Coherence-Oriented Contrastive Learning of Musical Audio Representations
Learning representations that capture how well musical stems belong together.
Audio ML +
Efficiency.
I’m Emilian. I work on speech enhancement, generative audio and efficient inference. My research includes an ICLR 2024 oral presentation, and I served as an Area Chair at ICASSP 2026. I’m currently interested in making audio models that run efficiently in highly constrained environments.
CURRENTLYLead Machine Learning Researcher
IRIS Audio Technologies / London, UK
IRIS Audio Technologies / London, UK
MSDM / ICLR 2024
01 / 05
RESEARCH FOCUS
Speech enhancement
Generative audio
Efficient inference
01 / Selected research
Ideas into sound.
A selection of collaborative work across audio generation, representation learning and faster inference.
5 SELECTED PAPERS
2025ICASSP
Naturalistic Music Decoding from EEG Data via Latent Diffusion Models
Exploring music reconstruction from non-invasive brain recordings with latent diffusion.
2024ICLRORAL
Multi-Source Diffusion Models for Simultaneous Music Generation and Separation
A shared generative model for synthesising music, separating sources and creating accompaniments.
2024ICASSP
SyncFusion: Multimodal Onset-synchronized Video-to-Audio Foley Synthesis
Generating controllable Foley sound that follows the timing of actions in video.
2023ACL
Accelerating Transformer Inference for Translation via Parallel Decoding
Parallel decoding algorithms that accelerate existing translation models without retraining.
* Equal contribution.
02 / Experience
Research, with
real-world intent.
Feb 2026 — Present
Lead Machine Learning Researcher
IRIS Audio Technologies · London
Aug 2024 — Jan 2026
Senior Machine Learning Researcher
IRIS Audio Technologies · London
Optimising models for real-time speech enhancement on embedded systems.
Mar — Jul 2024
Research Fellow
Ca’ Foscari University of Venice · Venice
Generative modelling for signal processing.
Nov 2023 — Feb 2024
AI Researcher
Sony CSL · Tokyo
Music decoding from EEG using latent diffusion models.
May — Sep 2023
Academic Visitor
Queen Mary University of London (C4DM) · London
Compositional music generation, singing voice separation and Foley synthesis.
Jun — Sep 2022
AI Researcher
Dolby Laboratories · Barcelona
Universal sound separation and efficient research workflows.
Earlier experience
Jun 2019 — May 2020
Junior Research Fellow
Sapienza University of Rome · Rome
Geometry processing.
Jun — Sep 2018
Software Developer
Babelscape · Rome
Porting the BabelNet semantic network from Java to Python.
03 / About
A little background.
My work sits between audio signal processing and machine learning: understanding sound, building generative models, and making them efficient enough to use.
I earned my Ph.D. in Computer Science at Sapienza University of Rome, advised by Prof. Emanuele Rodolà, graduating with honours in 2024. My research has taken me from Rome and Venice to Barcelona, Tokyo and London.
Romanian & Italian native · English & Spanish fluent
EDUCATION / SAPIENZA UNIVERSITY OF ROME
- Ph.D. in Computer Science
- 2020–2024 · Excellent with honours
- M.Sc. in Artificial Intelligence & Robotics
- 2018–2020 · 110/110 with honours
- B.Sc. in Computer Science
- 2014–2018 · 110/110 with honours
COMMUNITY & RECOGNITION
- ICASSP 2026Area Chair & Session Chair
- NeurIPS 2025Top Reviewer
- Research communityReviewing for ICLR, NeurIPS, ICASSP, AAAI and Interspeech
- Mentorship & teachingICASSP MiME mentor; teaching assistance at Sapienza
04 / Get in touch
Let’s talk
audio & ideas.
For research conversations, collaborations and speaking enquiries.