Jesse Clark

About

I’m a researcher and co-founder of Celeris and Marqo .

My background is in machine learning, physics and inverse problems. I have worked on language model efficiency, multimodal search, robot perception and continual learning, and computational imaging of nanoscale dynamics.

Google Scholar GitHub LinkedIn

Experience

Celeris & Marqo

Co-founder

At Celeris, I work on fast language and decision models, including model architectures and inference systems. At Marqo, I work on multimodal learning, search and retrieval.

Amazon Robotics

Machine learning lead

I worked on perception and continual learning, designing and leading the implementation of the Janus framework and the deep learning models for Robin’s perception system.

Stanford & UCL

Physics & inverse problems

I developed lensless imaging methods using non-convex optimisation to reconstruct nanoscale structures and dynamics. At Stanford, I held a Volkswagen Foundation Paul-Ewald Fellowship.

Research interests

Parallel language generation
Optimisation and model efficiency
Model architectures and inference systems
Multimodal learning and retrieval