Scholar
Filippos Kokkinos
Google Scholar ID: uuXQjUIAAAAJ
Research Scientist, Meta Superintelligence Lab
Citations & Impact
Citations
11,765
H-index
16
i10-index
20
Publications
20
Co-authors
0
Publications
Resume (English only)
Academic Achievements
- Produced nine peer-reviewed publications (five first-author at CVPR, ECCV, NeurIPS, and related venues). Contributed to projects such as 'Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training' and 'WildCAT3D: Appearance-Aware Multi-View Diffusion in the Wild'.
Research Experience
- Currently leading a vision–language reasoning team at Meta’s Super Intelligence Lab. Previously developed a complete text-to-3D generation stack that combines diffusion with neural rendering techniques—NeRFs, Gaussian splats, and mesh optimization—to create high-fidelity, editable 3D assets from natural-language prompts.
Education
- Ph.D. from UCL, supervised by Iasonas Kokkinos; published three NLP papers as an undergraduate at NTUA and won an NAACL/SemEval sentiment competition.
Background
- Multimodal GenAI researcher at Meta. Research interests include multimodal and generative models, reinforcement learning, text-to-3D diffusion, neural rendering (NeRF, Gaussians), and 3D geometry & reconstruction.
Miscellany
- Open to collaboration on foundational multimodal research. Believes that the future of AI is inherently multimodal, with text being only the beginning.
Co-authors