Alkis Koudounas
PhD Candidate & ML Researcher | Turin, IT
I am a half Greek-half Italian PhD candidate at Politecnico di Torino, Italy, specializing in speech processing and responsible AI. My current research focuses on aligning speech language models with human feedback. I love exploring problems and envisioning unseen solutions.
Selected Publications
voc2vec: A Foundation Model for Non-Verbal Vocalization
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025
A Contrastive Learning Approach to Mitigate Bias in Speech Models
Interspeech (Best Student Paper Award), 2024
Open-Source Projects
voc2vec suite
📥
6,594 downloads
voc2vec is the first universal representation model designed for non-verbal human vocalization tasks.
AudioSet-pretrained suite
📥
19,589 downloads
SSL models (HuBERT, wav2vec 2.0) pre-trained on AudioSet for general-purpose audio representation learning.
DeepDialogue
📥
38,464 downloads
DeepDialogue-orpheus is a large-scale multimodal dataset containing 40,150 high-quality multi-turn dialogues spanning 41 domains and incorporating 20 distinct emotions with coherent emotional progressions.
Experience
Applied Research Intern
Amazon AGI - Speech & Audio Foundation, Sunnyvale, CA
Aug - Nov 2025
Worked on reward modeling for speech language models. The model is currently in production.
Applied Research Intern
Amazon AGI - Speech Understanding, Aachen, DE
Jun - Dec 2024
Worked on reward and confidence modeling for automatic speech recognition and speech translation systems.
Research Intern
Thales Alenia Space, Turin, IT
May - Nov 2021
Worked on object detection and 6DoF pose estimation with ToF cameras.
Research Intern
Tokyo University Agriculture and Technology, Tokyo, JP
Mar - May 2019
Worked on gradient-based learning methods extended to manifolds.
Education
PhD in Computer Engineering
Politecnico di Torino, Turin, IT
2022 - Present
Thesis: Toward Robust, Responsible and Trustworthy Speech Foundation Models
MSc in Computer Engineering
Politecnico di Torino, Turin, IT
2019 - 2021
Thesis: Object detection and 6DoF pose estimation with ToF cameras
BSc in Computer Engineering
Università Politecnica delle Marche, Ancona, IT
2016 - 2019
Thesis: Gradient-based learning methods extended to manifolds
Other
Teaching
Teaching Assistant
Politecnico di Torino, 2023 - Present
Assisted in teaching graduate courses on data science and machine learning fundamentals.
Conducted lab sessions and graded assignments.
Guest Lecturer - Deep Learning for Speech Processing
Politecnico di Torino, 2023 - Present
Delivered lectures on transformer architectures and their applications in speech, audio and music processing and generation.
Awards & Honors
Outstanding PhD Award
Politecnico di Torino, 2025
Best Student Paper Award
Interspeech, 2024
Travel Grant Recipient
Interspeech, ICASSP, KDD 2024
Service
Organizer
SPADE Workshop at ICASSP, 2024
Online Chair & Head of Volunteers
ECML-PKDD, 2023
Reviewer
Interspeech, ICASSP, SLT/ASRU, KDD
Volunteering & Extracurriculars
Tutor & Supervisor
MALTO student team, 2023 - Present
AYA Italian Language Ambassador
Cohere for AI, 2024