Patrick Marmaroli
Freelance
Speech & Audio ML Engineer — signal processing to production software
Fully remote from the Geneva area (Switzerland / France)
Freelance
Speech & Audio ML Engineer — signal processing to production software
Fully remote from the Geneva area (Switzerland / France)
I build speech and audio AI that works outside the lab, from signal processing and model fine-tuning to the production software around it. My focus is demanding domains where errors are costly: clinical voice analysis, real-time communication quality, regulated multi-party workflows. Since August 2026, I work full-time as an independent freelancer through Hidacs Sàrl, fully remote from the Geneva area, and I co-founded Vocametrix, a speech analysis platform for speech-language pathologists. I run an AI-assisted development workflow that lets me own a project end to end — architecture, implementation, testing, and deployment. Previously Senior Audio Engineer at Microsoft (Teams audio quality), 2022–2026. PhD from EPFL, 25+ publications (7 journal articles, 740+ citations), 2 granted US patents in voice capture and hearing assistance.
Vocametrix OÜ · Tallinn, Estonia
Voice-controlled games and audio analysis tools for speech therapy. Web platform combining pronunciation assessment, pitch tracking, and gamified exercises, backed by a speech processing API suite serving both the platform and external developers.
Hidacs Sàrl · Switzerland · Remote
Freelance missions in speech & audio ML, signal processing and production software — feasibility studies, proofs of concept, model improvement and end-to-end builds. Selected missions: business platform suite (ERP, e-signature, timesheets, single sign-on) for a multi-company railway infrastructure alliance; phoneme recognition models for clinical speech therapy; core technology of an acoustic liveness detection solution for a biometric startup later acquired by Onfido; projects in biomedical, luxury watchmaking, security and environmental acoustics.
Microsoft · Tallinn, Estonia
Sound quality engineering for Microsoft Teams — audio processing pipelines, quality metrics, and signal analysis at scale.
Development of an anti-spoofing technology using ultrasound for identity verification liveness detection.
EARSENS project — privacy-preserving indoor activity detection using electrodynamic loudspeakers and AI, without microphones. Co-authored publication in Applied Acoustics (January 2023).
HED Technologies · Geneva, Switzerland
Designed and implemented signal processing algorithms (echo cancellation, noise reduction, voice command recognition) and contributed to headphone acoustic design. Led R&D efforts growing the Audio Department from scratch to a 4-engineer team, steering algorithm development from prototype to validated product.
PRONA SA · Yverdon, Switzerland
Developed custom Matlab tools and data processing pipelines for building and environmental acoustic analysis. Conducted normative acoustic measurements and managed multidisciplinary acoustic projects for industrial, institutional, and private clients.
EPFL · Lausanne, Switzerland
Designed audio signal processing algorithms and conducted laboratory and field experiments within European research projects. Collaborated with industrial partners including Phonak (hearing aids), FIAT, and ARMASUISSE. Supervised interns and PhD students.
Swiss Federal Institute of Technology (EPFL) · Switzerland
Bimodal sound source tracking applied to road traffic monitoring.
Université du Sud Toulon-Var · France
Signal processing, psychoacoustics, and trajectography.
ISEN Toulon · France
Lycée Lachenal · Argonay, France