Rhodes Scholar · University of Oxford
Firas Darwish
I am a Rhodes Scholar and PhD student in
Statistics and Machine Learning
(StatML)
at the University of Oxford.
I work on learning-based methods that enable robots to perceive, act, and
generalise across tasks and environments.
I also teach on executive education AI programmes for the governments of the UAE and South Korea. Previously, I graduated summa cum laude in Computer Engineering from NYU Abu Dhabi, where I worked with Farah Shamout. I also worked on AI Governance in e&.
Selected Work
All publications →
Projects
All projects →realtime-rl-code
Code release. Decision-making under real-time inference constraints.
semantic3d
End-to-end pipeline for reconstructing 3D scenes from RGB video, detecting semantically labeled 3D objects, and querying the resulting scene through an LLM interface.
ConvKAN3D
Python package for 3D convolutional layers using Kolmogorov-Arnold Networks (KANs).
Writing
All writing →Presentation
Motion Capture is Not the Target Domain Talk
This is the talk I gave presenting my work "Motion Capture is Not the Target Domain: Scaling Synthetic Data for Learning Motion Representations" to the Oxford Wearables Group.
Presentation
An Innocent Mistake is All it Takes
This is the keynote talk I gave during the Claude Buildathon at the University of Oxford on building safely with AI.
Presentation
World Models: A Path Towards Machine Intelligence?
This is a presentation on a subset of recent World Modelling literature as well as a discussion on some of the limitations of current World Modelling work.