Théo Morales

About me

I am a Machine Learning engineer at SPECS (Snap Inc.) working on Computer Vision problems for the next-gen Augmented Reality (AR) glasses.

I was previously a postdoc researcher at Trinity College Dublin under Prof. Binh-Son Hua. My research was focused on 4D reconstruction using dynamic Gaussian Splatting methods.

I got my PhD at Trinity College Dublin in 2025, supervised by Prof. Gerry Lacey. During my PhD, my research focused on hand-object interaction representation, 3D computer vision, and generative models.

Research interests

During my postdoc, I spent some time working on Gaussian Splatting for dynamic content and synthetic data generation for dynamic radiance fields. I benchmarked many SOTA methods and looked into their limitations. I published research that addresses fundamental limitations of the Gaussian Splatting framework, showing benefits in both static and dynamic scenes.

During my Ph.D., my research interests revolved around representation learning, meta-learning, hand-object interaction priors, generative modelling, etc. While I expressed a deep interest for the theory of deep learning and the family of latent variable models, my research was on the application of deep learning to 3D problems in the area of hand-object interaction.

Since I started my postdoc in 2024, I developed an interest for computer graphics, radiance fields, and 3D/4D reconstruction. I realized that my skills lie at the intersection between 3D computer vision and graphics.

My current research interests are:

Scene understanding and reconstruction
3D/4D Gaussian Splatting
Spatio-temporal modelling of scene dynamics
Radiance Fields

Publications

Related experience

Reviewer: CVPR, NeurIPS.
Teaching:
- Introduction to Linear Algebra for Engineering (Fuzhou University, China)
- Computer Graphics (Fuzhou University, China)
- Mixed Reality (Trinity College Dublin, Ireland)