Music, machine learning, and interaction

ANIMA — AI-based Interactive Microtonal Compositional Assistant

Marie Skłodowska-Curie Postdoctoral Fellowship · Music Technology Group, Universitat Pompeu Fabra, Barcelona · September 2025 – 2027

AI-driven tools for microtonal composition that keep the composer's creative agency at the centre. ANIMA builds a corpus and computational models of harmony beyond twelve-tone equal temperament (31-TET and 53-TET): the Harmonic Eigenspace, a psychoacoustic model of tetrad dissonance that turns the chord space into a navigable 3D topology of consonance; and GPT-style language models trained on microtonal harmony with a hierarchical tokenisation for 53-TET MPE MIDI. Supervised by Sergi Jordà, with a secondment at the Algomus team (CRIStAL, CNRS – Université de Lille) under Mathieu Giraud, and in collaboration with Ken Déguernel and Bob L. T. Sturm.

ANIMA project website →

Modelling harmony with transformers — MUSAiC

Postdoctoral researcher · Speech, Music and Hearing, KTH Royal Institute of Technology, Stockholm · 2022 – 2025

Three years in the ERC-funded MUSAiC project directed by Bob L. T. Sturm, modelling harmonic progressions with transformer language models: The Chordinator (token strategies for chord-progression generation) and ChromaFlow (voicing-aware generation of harmonic progressions). Also analysed the musical form and structure of large collections of AI-generated music from Udio and Suno, compared with human-made pop music. Preceded by a one-year Margarita Salas postdoctoral fellowship at UPF (2021).

Publications →

Musical gesture recognition — TELMI

PhD · Machine Learning Lab, Music Technology Group, Universitat Pompeu Fabra · 2016 – 2020

Machine-learning and deep-learning approaches to recognising and modelling instrumental gestures — bowing technique, fingering, expertise level — from IMU, motion-capture and EMG sensors, within TELMI (Technology Enhanced Learning of Musical Instrument Performance, H2020), supervised by Rafael Ramírez.

BrainX3 — large-scale neuronal data exploration

Research assistant · SPECS, Universitat Pompeu Fabra · 2014 – 2016

Software development for BrainX3, an interactive visualisation of the human connectome — a model of zone-to-zone brain connectivity — projected in the eXperience Induction Machine and used for neurosurgical planning and connectomic research (CEEDs project).

Project page →

Teaching

UPF · LCI Barcelona · IAAC · Elisava · Laboral Art Centre · Università Sapienza

Graduate courses in advanced interaction design, data visualisation and sound & music computing; undergraduate courses in C, Python, data structures and algorithms, numerical methods and computer organisation at UPF; supervision of master and bachelor theses. Workshops on programming and interaction design for more than fifteen years.