John Co-reyes
Member Of Technical Staff at Reflection AI
San Francisco, California, United States
Summary
👤
Senior🎓
Top SchoolJohn Co-reyes is a Member of Technical Staff at Reflection AI with 12 years of experience building frontier open-source models and applied reinforcement learning systems. Previously a research scientist at Google DeepMind, he led RL and self-improvement efforts for Gemini that produced state-of-the-art gains in coding, reasoning, and model revision workflows. He completed a PhD at UC Berkeley under Sergey Levine, publishing first- and second-author papers at ICLR, ICML, NeurIPS, and CoRL on model-based RL, meta-learning, and latent dynamics. John combines deep research instincts with engineering impact—shipping algorithmic changes into large LLM recipes, scaling synthetic data for code, and contributing practical ML tooling such as video loaders for Intel’s Nervana neon. He focuses on algorithms that let agents autonomously acquire diverse skills across robotics, vision, and NLP, and has a track record of turning novel RL ideas into production-going systems. Based in San Francisco, he blends rigorous academic pedigree with hands-on productization of cutting-edge AI.
12 years of coding experience
7 years of employment as a software developer
Doctor of Philosophy - PhD, Artificial Intelligence, Doctor of Philosophy - PhD, Artificial Intelligence at University of California, BerkeleyWilton High SchoolCalifornia Institute of Technology
English, Spanish
Github Skills (7)
video-processing10
computer-vision10
ffmpeg10
cprogramming-language9
c-language9
deep-learning8
python7
Programming languages (2)
Github contributions (5)
NervanaSystems/neon
Apr 2016 - May 2016
Intel® Nervana™ reference deep learning framework committed to best performance on all hardware
Role in this project:
ML Engineer
ML EngineerContributions:13 commits, 7 comments in 1 month
Contributions summary:John primarily focused on the video data loading and processing components within the Intel Nervana deep learning framework. They addressed issues in the image loader, specifically related to video pixel format warnings, and made changes to the `video.hpp` and `image.hpp` files. Furthermore, the user contributed to the video C3D demo by integrating the video loader and making minor adjustments to improve functionality.
deep-learningbest-performanceintelmachine-learningperformance
jcoreyes/OP3
Nov 2019 - Dec 2019
Contributions:3 releases, 6 commits, 4 pushes in 23 days
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.