CV

Contact Information

Name Gerardo Roa-Dabike
Professional Title Machine Learning Researcher -- Speech & Audio
Website https://www.gerardoroadabike.com/

Professional Summary

Machine learning researcher specialising in deep learning for speech and audio, including speech separation, speech enhancement, and noise cancellation. PhD in Computer Science with a focus on speech and audio processing, and a track record of designing, implementing, and evaluating ML systems – from CNN and transformer-based models to full evaluation frameworks and benchmarks – for real-world audio applications, published in peer-reviewed venues. Experienced in translating research into deployable systems through cross-functional and industry collaboration.

Experience

  • 2022 - present

    Uk

    Cadenza Project - Research Associate - ML Challenge Lead
    University of Sheffield, UK & University of Salford, UK
    Led the design and delivery of international machine learning challenges on audio enhancement and speech separation for hearing-aid users.
    • Led the design and delivery of 5 international ML challenges on audio enhancement and speech separation for hearing-aid users.
    • Designed and implemented baseline ML systems for speech separation and enhancement using CNN- and transformer-based architectures.
    • Benchmarked systems against real-time inference and latency constraints alongside challenge submissions.
    • Defined data strategies and rigorous evaluation frameworks for large-scale research initiatives across academic and industry partners.
  • 2022 - 2024

    UK

    COVID-OSA Project - Research Associate - Data Analytics Lead
    University of Sheffield
    Led data collection and analytical activities for a study on the relationship between obstructive sleep apnoea and COVID-19 outcomes.
    • Applied signal processing and statistical analysis to clinical and physiological data.
    • Curated a clinical dataset from 120 long-COVID and OSA patients that remains in active use by other researchers.
  • 2004 - 2017

    Chile

    Deputy Manager, Business Intelligence
    LarrainVial S.A.
    Led business intelligence operations and data infrastructure at a leading Chilean financial services firm.
    • Consolidated multiple product databases into a unified analytics platform.
    • Designed KPIs to benchmark 50–100 financial advisors.
    • Directed cross-departmental innovation projects.

Education

  • 2018 - 2022

    Sheffield, UK

    PhD
    University of Sheffield
    Computer Science
    • Thesis on deep learning for speech and audio processing, including speech separation and enhancement.
    • Focused on evaluation methodology for adapting spoken speech technologies to singing.
  • 2015 - 2016

    Sheffield, UK

    MSc
    University of Sheffield
    Advanced Computer Science
  • 1998 - 2003

    Chile

    Bachelor of Engineering
    Universidad Diego Portales
    Informatics Engineering

Publications

Projects

  • CAD1

    Machine learning challenge to improve music listening for people with hearing impairment.

    • Co-developed the first Cadenza Challenge with headphone and in-car listening scenarios.
    • Contributed to baselines, evaluation methodology, and open benchmark data release.
  • ICASSP 2024

    Personalised remixing of music for hearing aid users, accounting for loudspeaker cross-talk.

    • Extended Cadenza to a more realistic loudspeaker listening scenario.
    • Supported evaluation using HAAQI on listener-personalised remixed stereo signals.
  • PhD Project

    Deep Learning Approaches for Automatic Sung Speech Recognition

    • Focused on adapting spoken speech technologies to sung speech.
    • Covered dataset creation, vocal source separation, acoustic modelling, and lyric transcription.

Awards

  • 2026
    The School Citizen Award
    School of Computer Science at the University of Sheffield

    Recognising an exceptional contribution to the School community, going beyond the expectations of the role, through dedication, kindness, and a collaborative spirit.

Skills

Audio & Speech ML (Advanced): Speech Separation, Speech Enhancement, Noise Cancellation, Audio Transformers, CNNs, RNNs, Real-time / Low-latency Audio ML, Signal Processing
Data & AI (Advanced): Machine Learning, Deep Learning, Predictive Analytics, LLMs, Evaluation Frameworks, Data Curation, Statistical Analysis
Technical (Advanced): Python, PyTorch, SQL, Data Infrastructure, Linux
Leadership (Advanced): Research Leadership, Cross-functional Collaboration, Stakeholder Management, Technical Writing & Publications

Interests

Research Interests: Speech and audio ML, hearing technologies, music information retrieval, real-time DSP systems