Homepage
Sewade Olaolu Ogun, PhD
Welcome to my homepage. I am currently building interesting AI products at GetVocal AI in Paris. I recently finished my PhD in Computer Science at Inria.
My research areas include:
- Speech language models
- Generative text-to-speech systems
- Automatic speech recognition systems
- Dataset curation and augmentation
- Large language models
View CV on LinkedIn
Latest News
- 01 Jun 2025: Our paper on NiajaVoices got accepted at Interspeech 2025. This paper introduces over 1800 hours of speech-text pairs for the three major languages in Nigeria.
- 23 Oct 2024: I will give a practical talk to community members of AI Saturdays Lagos on the topic, “Implementation and evaluation of a research paper”. Slides will be available after the talk.
- 10 Oct 2024: I have successfull defended my PhD thesis titled “Generating diverse synthetic dataset for ASR training data augmentation”. Thank you to everyone who have supported me on this journey.
- 09 July 2024: I will be submitting my thesis at the end of July. My thesis defense is slated for October, therefore, I am on the lookout for new and interesting positions on deep-learning in research and in the industry.
- 04 June 2024: Two papers were accepted at Interspeech 2024. The first paper is in on improving NER for accented speech while the other paper is on TTS for African-accentend English. I will be presenting the two papers alongside my colleagues in Kos, Greece.
- 29 Nov 2023: I attended the Rencontres des Jeunes Chercheurs en Parole RJCP 2023 workshop in Grenoble. I had the opportunity to present one of my papers at the event.
- 20 Aug 2023: I will be attending Interspeech 2023 in Dublin. I will be presenting my work on improving the diversity and naturalness of TTS systems.
- 16 Jun 2023: I attended the vivatech conference, an annual technology conference, dedicated to innovation and startups, held in Paris. Thanks to INRIA innovation lab for sponsoring my trip.
- 05 Jan 2023: I will be presenting my work at Speech and Language Processing (SLT) workshop holding in Qatar. I will be presenting my work on curating high-quality datasets from crowdsourced datasets for TTS training.
- 02 Oct 2022: I implemented some iterative phase retrieval algorithms that can be used as vocoders for TTS.
- 20 Apr 2022: I will be attending the DeepLearn summer school in Guimarães, Portugal.
- 01 Aug 2021: I will be starting my PhD at Inria-Nancy and Vivoka in October. I will be working on builing an Automatic Speech recognition with limited data for Embedded Systems.
- 12 June 2021: We (a group of 8 research students) implemented the SAINT paper in the past week. It was so good working as a team to implement every detail in the paper
- 29 May 2021: I am volunteering with TREND IN AFRICA, a charity supporting scientific capacity building across Africa, to teach the Python programming language to beginners
- 24 Apr 2021: New blog post on Consider using UAR instead of Accuracy for Classification tasks
- 15 Apr 2021: I will be volunteering at the International Conference for Learning and Representations, ICLR 2021
- 01 Mar 2021: I started a Machine Learning Research Internship with ubenwa.ai. I will be working on predicting infant asphyxia through machine learning
- 26 Feb 2021: I participated in the qualifying rounds of Google Hash Code 2021 online competition, our team, Team OKAY was ranked amomg the top 25 percent on the leaderboard
- 26 Jan 2021: New blog post on How to create a speech dataset for ASR, TTS, and other speech tasks targetted at the speech processing community
- 01 Dec 2020: I will be a teaching assistant for the African Masters in Machine Intelligence programme at AIMS Senegal
- 01 Sep 2020: Finished cousework for my masters programme at AIMS AMMI. Currently looking for internship postions in Natural Language Processing or Speech Recognition
- 27 Aug, 2020: Teaching at a 2-day Python workshop for beginners alongside colleagues, with over 100 partipants. Workshop materials can be found in this github repo
Latest Articles
- Consider using UAR instead of Accuracy for Imbalanced Classification tasks
- How to create a speech dataset for ASR, TTS, and other speech tasks
- Breaking down the CTC Loss
- From GRU to Transformer
- K Nearest Neighbor as a Neural Network
- Making Efficient Neural Networks
For more articles, visit my blog.
Talks and Presentations
- Implementation and evaluation of a research paper, Talk at AI Saturdays research community
- Generating diverse synthetic dataset for ASR training data augmentation, PhD thesis presentation, Nancy
- 1000 African Voices: Advancing inclusive multi-speaker multi-accent speech synthesis, Poster presentation, InterSpeech 2024
- Stochastic Pitch Prediction Improves the Diversity and Naturalness of Speech in Glow-TTS, Poster presentation, InterSpeech 2023
- Data augmentation for speech processing tasks, Presentation for the speech corpus course, Masters TAL, University of Lorraine
- Can we use Common Voice to train a Multi-Speaker TTS system?, Poster presentation, SLT 2023, Doha
- Adversarial Examples, African Institute for Mathematical Sciences, Ghana.
- Chance Constrained Optimization, Convex Optimization Course, AIMS
Contacts
- Email: sogun [at] aimsammi [dot] org