What does it really mean for an audio model to be robust?
Short, noisy, reverberant, filtered audio; real hardware.
Gabriel Bibbó
Audio ML researcher and engineer
Audio ML · DSP · machine listening · acoustic privacy · real-time audio · MIR
Engineering since 2016Audio research since 202213 publications & worksEU work authorization
Updated 6 Sep 2026
Working questions
How much useful acoustic information can we retain without retaining speech?
Privacy-preserving datasets and release pipelines.
How do we make machine listening useful to musicians?
Harmonic mixing, library organisation, real-time analysis.
Selected work
Robustness, privacy, machine listening, and music technology.
Recovery Gain in Pruned Text-to-Audio Diffusion
Recovery fine-tuning can make a heavily pruned text-to-audio model look largely restored when it is evaluated under conditions similar to those used for fine-tuning. But that apparent recovery weakens sharply for shorter clips or out-of-domain prompts. The result shows why compressed generative models should be evaluated across real operating conditions, not at a single benchmark setting.
AudioLDM · Diffusion · Structured pruning · CLAP · PyTorch
Audio-Language Models for Voice Activity Detection
This research evaluates how audio-language models detect speech when the audio is short, noisy, reverberant, or filtered. It compares Qwen2-Audio-7B, Qwen2-Audio-7B with LoRA, Qwen3-Omni-30B, and Silero VAD. The best result came from Qwen2-Audio-7B with LoRA and OPRO-Template: 93.3% balanced accuracy on 21,340 degraded clips.
Qwen · LoRA · OPRO · Silero · PyTorch
Sounds of Home Dataset
Sounds of Home is a residential audio dataset for sound event detection. It contains 1,344 one-hour recordings collected from 8 participants in Belgium, using AudioMoth recorders placed in living rooms and kitchens. Speech was removed before release, and PANNs predictions were provided for the audio frames.
SED · AudioMoth · PANNs · Datasets
Speech Removal Framework
A framework for removing speech from audio recordings before they are shared or published. The system supports privacy-preserving release workflows while retaining non-speech acoustic information for sound event detection research.
PANNs · AST · Silero VAD · WebRTC VAD · Audio privacy
Harmonic EDM Mixing Compatibility
This music analysis system estimates how well two EDM tracks mix harmonically. It analyzes tracks, computes chroma features, converts them into Tonal Interval Vectors, compares harmonic compatibility, and suggests pitch shifts that can improve a mix. The work began as an MSc thesis and later became an ICWE 2022 publication.
Chroma · TIV · Essentia · librosa · EDM
ASR Enhancement Platform
An end-to-end MVP for comparing raw transcription with enhance-and-transcribe on pre-recorded audio. The backend persists jobs, audio artifacts, transcripts, and provider payloads, with FastAPI, Celery, PostgreSQL, Redis, MinIO, Docker Compose, metrics, tracing, Grafana, and CI. It is a reproducible engineering prototype, not a production-hardened service.
FastAPI · Celery · Docker · PostgreSQL · Redis
Traktor ML
Traktor ML turns a local Techno and Tech House library into Traktor-ready playlists. The pipeline extracts MERT embeddings, separates stems with Demucs, reads BPM and key metadata with Essentia, clusters similar tracks, orders them for smoother transitions, and exports M3U playlists. The current V4 run processed 239 tracks and exported 14 playlists.
MERT · Demucs · Essentia · HDBSCAN · UMAP · Streamlit
Experience
Jun.2026-Present
ML/DSP Engineer
Edge Audio Labs · Montevideo, Uruguay · Hybrid
- Applied machine learning, digital signal processing, testing, and perceptual evaluation across two confidential audio product lines, without disclosing client or project identities.
- Designed and delivered a rendering-side feature that maps score dynamics to model-level timbral expression rather than post-render gain alone, after reverse-engineering the end-to-end audio pipeline and identifying a hidden control-path failure.
- Built measurement and listening-test tooling covering approximately 580 renders and a 48-clip blind evaluation, then delivered the feature server-side without retraining the model.
- Built a headless C++ evaluation pipeline for real-time note and onset detection, from WAV and MIDI inputs through the production DSP to JSON metrics, regression tests, and adversarial canary cases.
- Found and corrected a systematic onset timing offset of approximately 104 ms, improved detector guard logic, and communicated results through technical documentation, pull requests, Jira, and client-facing presentations.
Dec.2025-Present
Visiting Researcher (collaboration)
University of Surrey · Remote · Research collaboration
- Preparing the manuscript ‘A Psychometric Evaluation of Audio-Language Models for Robust Voice Activity Detection’ for Elsevier Computer Speech & Language with Mark D. Plumbley and Simone Spagnol.
- Co-authoring work with Arshdeep Singh and Mark D. Plumbley on privacy-preserving audio and machine listening.
Nov.2022-Nov.2025
Research Engineer in Sound Sensing
University of Surrey · Guildford, UK · Research engineering
- Developed end-to-end audio ML systems for real-world smart environments, covering data preparation, model evaluation, prototype deployment, open-source releases, demos, datasets, and technical documentation.
- Built privacy-preserving SED pipelines for sensitive in-home recordings, including a 197 GB residential audio dataset, speech-removal workflows, and reproducible evaluation resources.
- Built an eight-model VAD benchmark on CHiME-Home and, separately, evaluated audio-language models under controlled duration, noise, reverberation, and spectral degradations.
- Deployed real-time CNN inference on Raspberry Pi, including latency, thermal, efficiency, and robustness evaluation for edge sound sensing.
- Published and presented research at ICASSP, IEEE WASPAA, CHiME Workshop, Inter-Noise, SMC, UKAI, UKIS, and AES. Supervised undergraduate and master’s projects.
Mar.2022-Nov.2022
Technical Support Engineer - Google Workspace
Webhelp · Barcelona, Spain · Enterprise support
- Tier 3 support for Google Workspace enterprise customers across APIs, OAuth, SAML/SSO, IAM, user provisioning, data migration, DNS/domain configuration, and security/compliance settings.
Nov.2021-Mar.2022
IT Auditor
KPMG · Barcelona, Spain · Technology audit
- Supported telecommunications companies and IT departments in technology audit engagements.
Aug.2016-Dec.2019
R&D Engineer
Ikatu · Montevideo, Uruguay · Embedded R&D
- Designed and shipped embedded C/C++ audio and IoT firmware for Bang & Olufsen home automation products, including low-level drivers, hardware integration, audio I/O, and Internet connectivity.
- Worked across requirements, architecture, implementation, testing, validation, and customer-facing documentation.
- Trained and onboarded incoming programmers in embedded development practices.
Apr.2016-Jul.2016
Engineering Intern
Ikatu · Montevideo, Uruguay · Engineering internship
- Developed and coordinated a complete home automation system project before transitioning into the R&D Engineer role.
Education
Sep.2020-Aug.2021
MSc Sound and Music Computing
Universitat Pompeu Fabra · Barcelona, Spain
60 ECTS · Computational models, audio engineering, perception, cognition, and interactive systems
Master thesis on harmonic compatibility for EDM mixing. Final thesis grade: 9/10.
Feb.2012-Dec.2017
BSc Electrical Engineering
Universidad de la República - Facultad de Ingeniería · Montevideo, Uruguay
Equivalent to 300 ECTS · Specialisation in electronics and signal processing
Bachelor thesis on autonomous mobile robots communicated by software-defined radio.
Mar.2002-Dec.2005
Formal musical training
School Nº265 ‘Virgilio Scarabelli Alberti’ · Montevideo, Uruguay
Musical language, choral singing, guitar, and instrumental ensembles
Publications
2025
-
Privacy for Audio AI: Risks, Challenges, and Emerging Solutions in the Era of Audio AI [Panel discussion]
Thomas Deacon; Jennifer Williams; Jason R. C. Nurse; Christopher Hicks; Gabriel Bibbó; Arshdeep Singh; Mark D. Plumbley
2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio
-
Speech Removal Framework for Privacy-preserving Audio Recordings
Gabriel Bibbó; Arshdeep Singh; Thomas Deacon; Mark D. Plumbley
2025 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), Tahoe City, CA, October 2025
-
Room Acoustics and Microphone Characteristics Show Systematic Impact on Sound Event Recognition
Gabriel Bibbó; Craig Cieciura; Mark D. Plumbley
Proceedings of the 54th International Congress and Exposition on Noise Control Engineering, São Paulo, Brazil, August 2025
-
Integrating IP broadcasting with audio tags: Workflow and challenges
Rhys Burchett-Vass; Arshdeep Singh; Gabriel Bibbó; Mark D. Plumbley
2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio
-
Soundscape Experience Mapping: A Deep Listening Approach for Eliciting Older Adults' Perceptions of Indoor Soundscapes
Thomas Deacon; Gabriel Bibbó; Arshdeep Singh; Mark D. Plumbley
Forum Acusticum / Euronoise 2025, Málaga, Spain, June 2025
-
Personalized Live Sound Recognition Using Efficient PANNs [Show and Tell]
Arshdeep Singh; Haohe Liu; Gabriel Bibbó; Thomas Deacon; Mark D. Plumbley
IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2025), Hyderabad, India, April 2025
2024
-
Environmental sound classification on an embedded hardware platform
Gabriel Bibbó; Arshdeep Singh; Mark D. Plumbley
INTER-NOISE and NOISE-CON Congress and Conference Proceedings, Nantes, France, August 2024
-
The Sounds of Home: A Speech-Removed Residential Audio Dataset for Sound Event Detection
Gabriel Bibbó; Thomas Deacon; Arshdeep Singh; Mark D. Plumbley
8th International Workshop on Speech Processing in Everyday Environments (CHiME 2024), Kos Island, Greece, September 2024
-
Soundscape Personalisation at Work: Designing AI-Enabled Sound Technologies for the Workplace
Thomas Deacon; Gabriel Bibbó; Arshdeep Singh; Mark D. Plumbley
International Conference on Sound and Music Computing (SMC 2024), Porto, Portugal, July 2024
2023
-
Recognise and Notify Sound Events Using a Raspberry PI Based Standalone Device [Demo]
Gabriel Bibbó; Arshdeep Singh; Mark D. Plumbley
IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA 2023), New York, U.S.A, October 2023
2022
-
A New Compatibility Measure for Harmonic EDM Mixing
Gabriel Bibbó; Ángel Faraldo
International Conference on Web Engineering (ICWE 2022), Bari, Italy, July 2022
2021
-
Towards a New Compatibility Measure for Harmonic EDM Mixing
Gabriel Bibbó
Master thesis, Universitat Pompeu Fabra, Barcelona, Spain, 2021. Supervisor: Ángel Faraldo.
2017
-
Autonomous Mobile Robots Communicated by Software Defined Radio
Gabriel Bibbó; Mariana Gelós; Martín Randall
Bachelor thesis, Universidad de la República, Montevideo, Uruguay, 2017. Supervisors: Pablo Belzarena and Federico Larroca.
Other projects
ALPACA
Software engineering prototype for algorithmic trading infrastructure, with market data ingestion, event processing, risk controls, historical simulation, persistence, API access, and monitoring. It has not been tested in production.
Python · Backtesting · Risk controls · Monitoring
Raspberry Pi Sound Event Recognition Demo
Raspberry Pi demo for real-time sound event recognition. The system runs pre-trained neural networks on a low-cost edge device, exposes a web interface, and can send email notifications when selected AudioSet events are detected.
Raspberry Pi · AudioSet · Real-time inference · Email notifications
3H-ATO
Mechanical tool designed during the pandemic to avoid touching shared surfaces directly.
Mechanical design · Prototyping · Product development
Automatic IoT Soap Dispenser
IoT handwashing device for industrial environments. The device used stainless steel, WiFi, cloud connectivity, IR/RFID sensors, and a 3-litre tank.
WiFi · IR/RFID · Cloud connectivity · Industrial hygiene
UyVoy Mobile App
Mobile app project for booking appointments and reducing crowding during the pandemic. Gabriel worked as product owner and project lead.
Product ownership · Civic tech · Project management
Courses & certifications
PRINCE2 Foundation in Project Management
Certificate GR656321771GB · Expires 03/2027
Project management techniques, risk mitigation, resource allocation, project planning, and monitoring.
Deep Learning
DeepLearning.AI · Coursera · Completed
CNNs, RNNs, transformers, TensorFlow, inductive transfer, and optimisation.
Machine Learning
Stanford University · Coursera · Completed
Supervised and unsupervised learning and best practices in machine learning.
Electronic Music Production
AURA · Montevideo, Uruguay · Graduated
Ableton and analogue instruments.
Private Guitar Lessons
Montevideo, Uruguay
Technical skills
Stack
Python · C/C++ · PyTorch · Hugging Face · PEFT · TorchAudio · librosa · Essentia · mido · scikit-learn · pandas · NumPy · SciPy · Flask · FastAPI · Streamlit · Docker · Git · Linux CLI · Bash · Slurm · Redis · Prometheus · Grafana · PostgreSQL · SQLite · MATLAB · Unreal Engine 5.4 · FMOD · VS Code
ML
CNNs · Transformers · Audio-Language Models · LoRA Fine-tuning · 4-bit Quantization · Supervised and Self-supervised Learning · Evaluation Pipelines · Statistical Testing · Edge Deployment
Audio
Sound Event Detection · Voice Activity Detection · Pitch and Onset Detection · Music Information Retrieval · Digital Signal Processing · Real-Time Audio · Perceptual Evaluation · DAWs · Ableton · DJing · Electronic Music Production
Practice
Reproducible ML pipelines · Automated Audio Testing · Dataset Curation · Open-Source Development · MLOps practices · AI-assisted Development · Technical Writing · Interdisciplinary Collaboration
Languages
Spanish Native · English C1 (CEFR) · Portuguese A2 (CEFR)
Memberships & funded research
IEEE Signal Processing Society (2025)
EPSRC AI for Sound · University of Surrey
Contact
Available for academic collaboration, consulting, contract research, and selected remote opportunities in audio ML, machine listening, DSP, and music technology.