Hi, I'm St John.
I'm a PhD researcher working on reinforcement learning, active inference, and causality. This is my personal site: the worked-out, the half-baked, and the opinions I couldn't keep to myself.
Featured
Where Should an Agent Direct Its Interoceptive Attention?
A fixed-budget attention mechanism points perceptual precision at the bodily need an agent currently believes is most urgent. In AffectWorld, this raises learning-phase survival from 19.9% to 41.4% at the same total budget. A companion to our SAB 2026 paper.
-
A Taste of Reinforcement Learning
A hands-on introduction to reinforcement learning — from the MDP framework and Bellman equation to Q-learning and beyond, with an interactive gridworld demo.
-
Predicting Protein Function with DeepChain
A walkthrough of how I created a DeepChain app for predicting Gene Ontology given only a protein sequence.
-
A Productive Study and Research Workflow with Notion
A guide on how to set up the ultimate research workflow with Notion.
-
Causal Reinforcement Learning: A Primer
A primer opening a series on causal reinforcement learning: Pearl's ladder of causation, and why RL agents need causal structure to reason beyond correlation.
-
World Models: Learning by Imagination
A walkthrough of the World Models paper, where an agent learns a compressed model of its environment and trains by dreaming inside it.
-
Learning with a Policy
An introduction to model-free deep RL: policy-gradient methods, the policy gradient theorem, and actor-critic.
