msc speech & language processing · university of edinburgh · 2025–2026
At Edinburgh, I studied machine learning across language, speech and vision. The MSc covered advanced NLP, computer vision, machine learning for signal processing and automatic speech recognition, with further work in retrieval, multilingual and multimodal models, distillation and Bayesian methods.
Before starting the MSc, I spent a year working for dataannotation.tech, producing and evaluating specialist training data for frontier language models across programming, mathematics, biology and scientific reasoning.
My dissertation, supervised by Catherine Lai and Sarenne Wallbridge, investigated auxiliary adaptation of full-duplex dialogue models for conversational coordination. I adapted PersonaPlex using dialogue-act and future-activity objectives, built reproducible training and evaluation pipelines, evaluated more than 1,000 paired conversations and ran a 21-participant listening study.
The results showed that improved performance on automatic timing measures did not necessarily translate into listener preference. That gap between benchmark performance and human judgement became one of the project’s central findings.