Open source · UNDP Mexico
Text mining for public policy
NLP techniques applied to Mexico's performance-evaluation system for public programs, developed with the UNDP Accelerator Lab. Read the publication.
“Creativity comes from applying things you learn in other fields to the field you work in.”
Hello, I'm Daniel Legorreta. I have spent over a decade working on data projects, taking on various roles such as data scientist, data engineer, ML engineer, MLOps engineer, and data analyst. Consequently, I do not limit myself to a single label, given the diversity of positions and projects I have handled. My experience ranges from building real-time forecasting systems to developing document data extraction solutions (including RAG systems) and AI agents. While I have a preference for certain technologies, I am passionate about solving complex problems, especially those that can have a positive impact on people's lives.
My core expertise includes:
- Advanced NLP and LLMs for large-scale information retrieval and text comparison.
- Scalable forecasting models in distributed environments.
- Anomaly detection systems.
- Computer vision and deep learning systems.
I constantly explore new techniques, algorithms, and technologies related to both model development and their deployment into production. I also enjoy sharing my knowledge through teaching and mentoring sessions whenever time permits.
Selected Work
A few public projects and publications. Much of my industry work is private, but more code lives on my GitHub.
Open source
RAD — robust anomaly detection
Robust PCA and robust deep autoencoders for anomaly detection in time series.
Publication · PLOS ONE
MicNet toolbox
Co-author of a toolbox for visualizing and unraveling microbial interaction networks (code, SparCC module).
Patent
WO2024178226
International patent application listed on WIPO PATENTSCOPE.
Blog
Thoughts, tutorials, and deep dives into AI and system design. Posts are published on Medium and appear here automatically.
Get In Touch
The best way to reach me is on LinkedIn.