Gustavo Pedro Ricou

MSc Statistics and Data Science (in progress), Trinity College Dublin

How models learn together. How measurements change conclusions.

I investigate how modelling and measurement choices shape quantitative conclusions. My current thesis at Trinity College Dublin studies transfer and gradient interference in simultaneous multi-task learning. Alongside research, I build backend integrations and live-data pipelines at BoscaSports, drawing on training in physics, statistics and cloud-native computing.

Available for PhD opportunities from July 2027

Gradient interference and transfer in multi-task learning

I compare networks that learn two tasks simultaneously from the same inputs, using a shared trunk and separate heads, with matched single-task networks. I investigate when sharing helps or hurts each task, how gradient interference relates to transfer, and whether transfer can be predicted before training.