My research focuses on end-to-end embodied AI for real-time situated interaction. I am particularly interested in online sensory integration and predictive representation learning for agents operating continuously in the real world: goal-directed, self-stabilizing dynamical models that support long-horizon, open-domain, multi-task interaction and no-reset task switching.
I try to approach these questions through a multidisciplinary lens, drawing from machine learning, cognitive science, and biologically inspired intelligence. Earlier projects have explored situated interaction, real-time task guidance, visual grounding in multimodal models, and human-action understanding.
Before joining Qualcomm, I was part of Twenty Billion Neurons (TwentyBN). I hold an M.Sc. in Artificial Intelligence from the University of Edinburgh and a B.Eng. in Aerospace Engineering from Ryerson University.
Recent Posts
All postsRisotto Stone
An AI cooking assistant for tracking recipe progress and giving timely cooking instructions.
Cross-Modal Generation and Modality-Driven Semantic Segmentation with Generative Adversarial Networks
M.Sc. dissertation work on audio-visual cross-modal generation, dataset curation, and GAN-based modality-driven segmentation.
Low-Resource Deep Reinforcement Learning for Four-Player Chess
A Machine Learning Practical project at the University of Edinburgh exploring AlphaZero-style self-play, MCTS, and reward shaping in four-player chess.
Academic Background
University of Edinburgh
M.Sc. in Artificial Intelligence, 2019
Ryerson University
B.Eng. in Aerospace Engineering with Honours, specializing in Avionics and Control Systems, 2016
Research Interests
I am interested in agents that learn by acting, maintain stable internal state over time, and adapt across long-horizon streams of changing tasks.
- Active vision for embodied agents
- Self-stabilizing dynamical systems
- Goal-driven reasoning, planning, and action with end-to-end neural architectures
- State refinement and exploration in continuous or discrete latent spaces
- Online sensory integration and multimodal future-predictive representations
- Interactive training for intrinsically agentic AI
Outside of Research
Beyond research, movement, language, and music recharge me: bouldering, wind-surfing, squash, skiing, photography, violin, singing, and the long project of becoming a hyper-polyglot. Meditation and introspection are a constant source of insight, and sometimes distraction. I find it hard not to be captivated by the moment-to-moment texture of experience, by cognitive biases as they surface, and by the limits and power of attention - though admittedly, hardest when it matters the most! My curiosity and bewilderment here shapes how I move through the world, how I relate to others, and why I keep returning to questions of mind, consciousness, and intelligence.