Mathias Gatti

ML Engineer · Audio, Video & Multimodal AI · Remote Contractor

Buenos Aires, Buenos Aires Province, Argentina

About

I'm a Machine Learning Engineer based in Buenos Aires with 8+ years building production AI systems for LatAm, US and European companies — remotely. My focus is multimodal AI: extracting insights from audio, video, image, and text at scale. I've worked across the stack — from fine-tuning state-of-the-art models to deploying them on cloud infrastructure — for startups in music tech, ad analytics, and speech synthesis. What I do: Audio & speech ML (TTS, voice conversion, music analysis) Video & image understanding (ad detection, content analysis) LLM integration and NLP pipelines End-to-end deployment (AWS, GCP, FastAPI, Docker) How I work: I take on part-time contractor engagements — typically async, delivery-focused, with no need for daily standups or fixed hours. I work well with small technical teams that need senior ML capacity without a full-time hire. Open to: contractor roles, part-time engagements, and technical consulting.

Experience

  • Machine Learning Engineer at OpenBrand
    Feb 2022 - Present · 4 yrs 5 mos

    I focus on automating the extraction of valuable insights from various media formats, including audio, text, video, and images. By developing advanced algorithms, I enable the efficient analysis of ads across podcasts, videos, and social networks, optimizing the way we derive information and generate value.

  • Music Software Engineer at mayk.it
    Jan 2023 - Apr 2024 · 1 yr 4 mos

    I specialized in cutting-edge singing synthesis and voice conversion, leveraging state-of-the-art models and implementing them on scalable platforms like covers.ai. My role involved bringing these advanced models into the real world to enhance the music production process.

  • Machine Learning Engineer at Storyteller.AI / FakeYou
    Mar 2022 - Oct 2022 · 8 mos

    During my tenure at FakeYou.com, I automated dataset creation tasks and pioneered a novel approach to fine-tuning text-to-speech models. By adapting the grapheme-to-phoneme system, I enabled seamless integration of these models into new languages, eliminating the need for training from scratch. This breakthrough significantly streamlined the development of multilingual speech synthesis systems

  • Lead Data Scientist at Unbias, Inc.
    Jul 2020 - Mar 2022 · 1 yr 9 mos

    Utilizing state-of-the-art Music Information Retrieval (MIR) models, I spearheaded the development of advanced machine learning pipelines for a cutting-edge music analytics platform. These pipelines enabled the extraction of valuable insights and analysis of music data, empowering users to make data-driven decisions in areas such as recommendation systems, genre classification, and music similarity.

  • Teacher at Instituto Humai
    Nov 2020 - Dec 2020 · 2 mos

    I worked preparing the material and teaching about web scraping and text generation with deep learning models. The classes are open source, they are available here: https://github.com/institutohumai/cursos-python