Fully Funded PhD position in Probabilistic ML for Audio
| Source: Mastodon | Original article
KU Leuven’s PSI division has opened a fully funded PhD slot dedicated to probabilistic machine learning for audio. The nine‑month‑long project will tackle how audio representations can be made robust across cultures and musical styles, while advancing sequence modelling, tokenisation, uncertainty quantification and information‑retrieval techniques for sound. Candidates must hold an MSc in electrical engineering, computer science or AI, demonstrate solid probability and coding skills, and submit a one‑page motivation letter outlining their experience with probabilistic ML.
The announcement arrives as probabilistic approaches gain traction in the broader AI ecosystem. Unlike deterministic deep nets, probabilistic models provide calibrated confidence scores, a feature increasingly vital for speech assistants, music recommendation engines and acoustic monitoring systems that must operate reliably in noisy, multilingual environments. By focusing on cross‑cultural audio representations, the research could reduce the bias that plagues many current speech‑recognition and music‑analysis tools, a concern echoed across the Nordic AI community.
The position also dovetails with recent interest in high‑performance, locally run AI pipelines – from the fully local OSINT agents built on Ollama and LangChain to GPU‑intensive probabilistic converters showcased earlier this month. Leuven’s emphasis on uncertainty and retrieval hints at future integrations with multimodal systems that blend sound, text and vision, a direction many Nordic startups are already exploring.
Watch for the application deadline (mid‑May) and the selection timeline, which will be announced on the university’s portal. Successful candidates are likely to present early results at conferences such as ICASSP or Interspeech, and may attract industry partnerships with audio‑tech firms seeking calibrated, culturally aware models. The PhD could become a conduit for Nordic researchers to collaborate on next‑generation audio AI that balances performance with trustworthy uncertainty estimates.
Sources
Back to AIPULSEN