Research Scientist Intern Current
Meta · Zurich
- Working on Meta's 3D GenAI team to advance 3D asset generation models, with an emphasis on efficient generation pipelines and downstream asset usability.
Kyle Fogarty
I'm a PhD student in Computer Science at the University of Cambridge, working across 3D computer graphics, computer vision, and machine learning.
Research Scientist Intern · Meta 3D GenAI About
Research
I work on computational methods that recover, represent, and generate the 3D world,
supervised by Prof. Cengiz Öztireli
in the Department of Computer Science & Technology at Cambridge. My research spans
neural surface reconstruction, point cloud generation, and 3D generative models, working
toward high-fidelity digital models of the visual world.
Alongside my PhD, I bring this research into practice through industry, building
production systems for 3D reconstruction and generation.
Focus
-
3D reconstruction
Recovering geometry and appearance from partial observations — neural implicit surfaces, point cloud generation, and shape recovery from images or scans.
-
Generative 3D
Diffusion, splatting, and latent models for synthesising and editing 3D content at scale.
-
Structured learning
Hard constraints, symmetries, and self-supervised priors that make 3D inference reliable in the wild.
Experience
Industry & applied research
Apr – Oct 2026
Jul 2024 – Feb 2026
Senior Deep Learning Research Scientist
- First ML research scientist hire at pre-seed; led model deployment end-to-end and scaled production ML systems that power $40M ARR.
- Built and shipped robust 3D reconstruction and geometry completion pipelines used by more than 30,000 users.
- Drove product roadmap and innovation with the CEO/CTO, mapping product needs to ML milestones.
Publications
Papers & preprints
2026
2025
An Information Theoretic Approach to Machine Unlearning
PSHead: 3D Head Reconstruction from a Single Image with Diffusion Prior and Self-Enhancement
Best Foot Forward: Robust Foot Reconstruction in-the-wild
Twist and Compute: The Cost of Pose in 3D Generative Diffusion
Education
Education
2022 – Present
PhD, Machine Learning
Thesis on Neural Geometry Processing — developing algorithms that embed structural priors into 3D ML tasks for improved robustness and generative fidelity.
2021 – 2022
MSc Robotics Distinction
University of Lincoln
Thesis on point cloud reconstruction, using neural basis functions for surface recovery from sparse 3D data.
2017 – 2021
MMath & Phys, Mathematics and Physics First Class
University of Manchester
Recovered strain in polycrystalline materials from neutron imaging spectra; work published in Applied Physics Letters.
News
Recent updates
- Apr 2026 Joined Meta’s 3D GenAI team in Zurich as a Research Scientist Intern.
- 2026 “Self-Supervised Implicit Attention Priors for Point Cloud Reconstruction” was presented at 3DV 2026.
- Nov 2025 “Twist and Compute” was accepted to the Principles of Generative Models workshop at NeurIPS 2025.
- Oct 2025 Two papers at the ICCV 2025 workshops: “Best Foot Forward” and “SYM3D”.
- 2025 “An Information Theoretic Approach to Machine Unlearning” was accepted to TMLR.
- Oct 2024 “FrePolad: Frequency-Rectified Point Latent Diffusion for Point Cloud Generation” was presented at ECCV 2024.
- 2024 Named a finalist for the Qualcomm Innovation Fellowship.
- Jan 2024 Joined the EnterpriseTech programme at Cambridge Judge Business School.
- Dec 2023 “Neural Fields with Hard Constraints of Arbitrary Differential Order” was presented at NeurIPS 2023.
- Oct 2022 Started my PhD in Computer Science at the University of Cambridge.