Senior Research Associate in Machine Learning for Speech Processing

Contract: Full-time (1.0 FTE), fixed-term [18 months]

The Project

Modern machine learning can predict vocal tract shapes from audio recordings of the voice with remarkable accuracy, but most of these models are black boxes. This Royal Society-funded project aims to crack open the black box and solve one of the most compelling challenges in speech science: understanding the mapping between vocal tract movements and the acoustic speech signal. Using state-of-the-art MRI recordings of the vocal tract during speech, we aim to develop machine learning approaches that don’t just predict acoustic output from articulatory configurations, but reveal why and how these mappings work. We need approaches that combine predictive power with scientific insight: models whose internal representations align with phonetic and physical knowledge. This requires hybrid machine learning (ML) approaches that integrate domain knowledge with data-driven learning, as well as explainable AI (xAI) techniques that make model behaviour transparent and scientifically meaningful. You will apply these approaches to a large database of real-time MRI and acoustic recordings of the vocal tract. Solving this problem will help to drive fundamental progress on critical applications, such as articulatory biofeedback for language learning and speech therapy.

Your Role

Working with Dr Sam Kirkham (Lancaster, Speech Science), Dr Anton Ragni (Sheffield, Computer Science) and Professor Aneta Stefanovska (Lancaster, Physics) you'll develop and validate interpretable ML approaches for modelling acoustic-articulatory relations using MRI vocal tract data. The position is available for 18 months from 1 July 2026 (start date negotiable).

Key objectives

• Develop hybrid ML architectures that incorporate phonetic and physical constraints.
• Apply and extend explainable AI techniques for speech production modelling.
• Validate model interpretations against established knowledge.
• Lead and/or contribute to publications at the intersection of speech science and machine learning.

This is a methodologically creative role with genuine intellectual ownership. You'll have access to rich MRI datasets and Lancaster's high-performance computing facilities.

Why This Role?

• Intellectual ownership – shape the methodology, not just implement it.
• Interdisciplinary impact – apply ML expertise to fundamental questions about human speech.
• Publication opportunities – first-author papers bridging AI and speech science.
• Flexibility – flexible working arrangements.
• Strong mentorship – work with researchers spanning speech production, computational modelling, machine learning, and biophysics.
• Career development – build expertise at the growing intersection of interpretable AI and speech processing.

Informal enquiries welcome: Dr. Sam Kirkham, s.kirkham@lancaster.ac.uk

Type / Role:

Subject Area(s):

Location(s):