Summary: Adapt language and vision models to
perform retrieval using image and text inputs.
Supervisors: Dr. Arnab Dhua, Dr. Xinliang Zhu, and
Michael Huang.
Applied Science Intern
Project: "Robust visual similarity learning”.
Summary: Derive a procedure to train robust deep
metric learning models on datasets with noisy
correspondences. Propose a procedure to perturb the labels
of benchmark that mimics real annotation errors.
Supervisors: Dr. Arnab Dhua, Dr. Xiaofan Lin, and
Dr. Muhammet Bastan.
Applied Science Intern
Project: "Contextual Position Bias Estimation for
Unbiased Learning to Rank”.
Summary: Propose a ranker-agnostic method to model
the probability of an item being seen in different
contexts, e.g., for different users, with a single
estimator. This can be used in LTR algorithms to
distinguish the absence of click due to irrelevance or
lack of examination, thus better matching the true user
preferences.
Supervisors: Dr. Matteo Ruffini and Dr. Giuseppe Di
Benedetto.
Engineering Intern, CTO
Project: ”Speech and Audio Style Transfer” -
Internship credited with Master Thesis.
Summary: This project introduced FastVC, a method
for fast (4x faster than real-time on CPU) many-to-many
non-parallel Voice Conversion (VC). The proposed approach
was accepted as a technical paper to the VC Challenge
2020, where FastVC outperformed the Challenge baselines in
quality results on the cross-lingual VC task.
Supervisors: Dr. Milos Cernak and Prof. Martin
Jaggi.
Summary: Proposed evolution strategies as zeroth
order optimization algorithms to find the worst affine
transforms for each input. The proposed method effectively
yields robust models and allows introducing non-parametric
adversarial perturbations.
Supervisors: Dr. Sebastian Stich, Maksym
Andriushchenko (PhD candidate at EPFL), and Prof. Martin
Jaggi.
Junior at Risk Advisory IT Department
Bachelor's internship.
Hardening security measures of IT systems at Banc
Sabadell.
Deep Learning Frameworks: PyTorch, TensorFlow, MXNet,
Keras
Languages
Catalan
Spanish
English
French
Honours
1st prize
SEAT Challenge - AI For Mobility & Driving Experience
1st prize
Bassoon category
Volunteering
Volunteer
Attence to virtual paper presentations to ensure the correct
functioning and the lack of violations of the code of conduct
as well as to stimulate the discussions.
Student Volunteer
Coordination of several keynotes and paper presentations.