Expert in Sequential Deep Learning

Core contributor to Genie 3, Veo 1, 2, 3, MoveNet, Youtube recommendations.
More than 10 years of experience in the field with applications to recommender systems, systematic trading, pose estimation and generative AI.

Placeholder

My bio in more details

Practical experience:

Currently I’m the CTO of Levenlight LLC where I lead research on Generative AI applied to embedded systems and robotics.

Before that I was one of the core contributors to Genie 3 and Veo 1, 2, 3 which I helped make real-time or near-real time. My career has focused a lot on neural network efficiency as in MoveNet which I helped make real-time by strongly improving the sequential filtering layer. I also worked on sequential recommendations for YouTube where I pioneered new classes of models to help people find videos they would love under tight computational constraints.

Aside from a background in AI, I also have practical hands on experience in quantitative finance, in particular statistical arbitrage and execution optimization.

Education:

Graduated from UC Berkeley in 2017 with a PhD in Computer Science specialized in AI.

Graduated from Imperial College in 2014 with an MSc in Distributed Computing Systems.

Graduated from Ecole Polytechnique in 2013 with an MSc in Probability and Finance.