Jose Javier is a research scientist focused on large language model pretraining and fine-tuning, bringing an MIT PhD and hands-on ML engineering experience at MosaicML and Databricks. In two years of industry work he has shipped optimizations in LLM training—improving in-context learning, compiling GLU layers, and integrating Triton RMSNorm—to boost performance in the widely used mosaicml/llm-foundry codebase. Based in San Francisco, he blends rigorous academic training with pragmatic engineering, tackling data loading, configuration, and logging issues that make research code production-ready. Co-author of The Missing Semester, he contributes to open-source tooling and emphasizes practical, reproducible ML workflows.
LLM training code for Databricks foundation models
Role in this project:
ML Engineer
Contributions:21 reviews, 24 PRs, 44 pushes in 1 year 10 months
Contributions summary:Jose contributed to the LLM-Foundry project by implementing and refining functionalities related to in-context learning and model optimization. Key contributions include modifying the `CodeEval` process, compiling the GLU layer for performance improvements, and integrating Triton RMSNorm for optimized normalization. The user also addressed issues related to configuration, data loading, and logging within the project.
Contributions:2 PRs, 62 pushes, 4 branches in 2 years 3 months
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.