Alkis Koudounas

Dottorando in Ingegneria Informatica E Dei Sistemi , 38^o ciclo (2022-2025)
Dipartimento di Automatica e Informatica (DAUIN)

Docente esterno e/o collaboratore didattico
Dipartimento di Automatica e Informatica (DAUIN)

alkis.koudounas@polito.it

Profilo

Dottorato di ricerca

Argomento di ricerca

Toward Robust, Responsible and Trustworthy Speech Foundation Models

Tutori

Interessi di ricerca

Data science, Computer vision and AI

Biografia

Alkis is a third-year Ph.D student at the Polytechnic University of Turin, Italy, and an applied research intern at Amazon AGI. His research focuses on speech, audio, and multimodal understanding, as well as robust, responsible, and trustworthy AI. His interest also lies in the development of resources for under-represented languages, and he is currently focusing on creating tools to help diagnose voice-related pathologies and improve the quality of life for those affected by them. He served as the Italian Language Ambassador for the AYA Cohere4AI project. His works have been published in top-tier speech and NLP conferences and journals, including ICASSP, ACL, IEEE/ACM TASLP, COLING, EACL, and Interspeech, where he also won the Best Student Paper Award (2024).

Personal website: https://koudounasalkis.github.io/

Premi e riconoscimenti

Applied Scientist Intern at Amazon AGI - Speech Recognition Team (2024)
Best Student Paper Award at ISCA Interspeech 2024 Conference (2024)
IEEE ICASSP 2024 Travel Grant Recipient (2024)
ACM KDD 2024 Travel Grant Recipient (2024)
ISCA Interspeech 2024 Travel Grant Recipient (2024)
Online Chair at ECML-PKDD 2023 (2023)
PhD's Pitch @ IEEE Polito Student Branch Title: “Subgroup Disparities in Speech Models: Detection and Mitigation” Short Abstract: The research addresses performance disparities in speech models across subgroups through in-processing (divergence-aware regularization, targeted augmentation, contrastive learning) and post-processing (targeted data acquisition) techniques, improving robustness for underrepresented demographics. (2025)

Mostra di piùMostra meno

Didattica

Insegnamenti

Corso di laurea magistrale

Data Science e Tecnologie per le Basi di Dati. A.A. 2024/25, INGEGNERIA INFORMATICA (COMPUTER ENGINEERING). Collaboratore del corso
Data science lab: process and methods. A.A. 2024/25, DATA SCIENCE AND ENGINEERING. Collaboratore del corso
Data Science e Tecnologie per le Basi di Dati. A.A. 2023/24, INGEGNERIA INFORMATICA (COMPUTER ENGINEERING). Collaboratore del corso
Data science lab: process and methods. A.A. 2023/24, DATA SCIENCE AND ENGINEERING. Collaboratore del corso
Data Science e Tecnologie per le Basi di Dati. A.A. 2022/23, INGEGNERIA INFORMATICA (COMPUTER ENGINEERING). Collaboratore del corso
Data Science e Tecnologie per le Basi di Dati. A.A. 2022/23, INGEGNERIA INFORMATICA (COMPUTER ENGINEERING). Collaboratore del corso
Data science lab: process and methods. A.A. 2022/23, DATA SCIENCE AND ENGINEERING. Collaboratore del corso

MostraNascondi A.A. passati

Ricerca

Gruppi di ricerca

DBDM - Database and Data Mining Group (DAUIN)

Pubblicazioni

Pubblicazioni più recenti Vedi tutte le pubblicazioni su Porto@Iris

Liu, Qingqing; Ciravegna, Gabriele; Koudounas, Alkis; Cerquitelli, Tania; Baralis, Elena (2025)
Multimodal Fusion Techniques to Enhance Voice Disorder Diagnoses. In: Workshops of the EDBT/ICDT 2025 Joint Conference, DARLI-AP Workshop, Barcelona (ESP), 25-28 March, 2025. ISSN 1613-0073
Contributo in Atti di Convegno (Proceeding)
Koudounas, Alkis; La Quatra, Moreno; Siniscalchi, Sabato Marco; Baralis, Elena (2025)
voc2vec: A Foundation Model for Non-Verbal Vocalization. In: 2025 IEEE International Conference on Acoustics, Speech and Signal Processing, Hyderabad (IND), 06-11 April 2025, pp. 1-5. ISBN: 979-8-3503-6874-1
Contributo in Atti di Convegno (Proceeding)
Koudounas, Alkis; Pastor, Eliana; Alfaro, Luca de; Baralis, Elena (2025)
Mitigating Subgroup Disparities in Speech Models: A Divergence-Aware Dual Strategy. In: IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, vol. 33, pp. 883-895. ISSN 2329-9290
Contributo su Rivista
Giobergia, Flavio; Koudounas, Alkis; Baralis, Elena (2024)
Large Language Models-aided Literature Reviews: A Study on Few-Shot Relevance Classification. In: 2024 IEEE 18th International Conference on Application of Information and Communication Technologies (AICT), Turin (ITA), 25-27 September 2024. ISBN: 979-8-3503-8753-7
Contributo in Atti di Convegno (Proceeding)
Koudounas, Alkis; Giobergia, Flavio (2024)
Houston we have a Divergence: A Subgroup Performance Analysis of ASR Models. In: 49th IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops, ICASSPW 2024, Seoul (KOR), 14-19 April 2024, pp. 812-813. ISBN: 979-8-3503-7451-3
Contributo in Atti di Convegno (Proceeding)