I am developing new machine learning methods to deepen human-machine collaboration by breaking
down structured tasks like programming into a conversation. My adviser is
Jeff Dalton at the
Glasgow Representation and Information Learning Lab (GRILL).
My fundamental motivation is to build agents capable of deep collaboration to achieve complex goals through conversation to enhance rather than substitute humans. In particular, AI partnerships to generate and refine program code.
Complex tasks are abundant. However, programming benefits from being fully digitized while still involving all essential aspects of complex problem solving. Its structured and brittle nature poses an additional challenge since we generally communicate through diffuse ideas rather than concrete implementations.
I graduated with First Class Honours in Computing Science where, in my last year, I focused on Algorithmics and Artificial Intelligence and defended my thesis on using deep recurrent networks for language conditioned code generation. Along the way I squeezed in two summer internships at Birdi, an AI Startup in satellite imagery where I worked in a great team led by Sivakumaran S. as a research scientist in computer vision helping push ML models to production. The research also included high performance data visualizations for customer click data leading to a novel sweep line pruning algorithm that was presented at UCL.