ABOUT · MIGUEL ESCALANTE

What happens after the model runs.

Projects rarely fail over two points of AUC. They fail when the human bottleneck downstream gets ignored. I learned that building crime prediction models for the Mexico City police, and it held through federal open data, earthquake response, pricing pipelines, and national healthcare.

Today I lead AI strategy and clinical AI systems for the Government of El Salvador. I studied Applied Mathematics (B.Sc.) and Data Science (M.Sc.) at ITAM, where I taught graduate courses for four years while working full-time. Part 1 and Part 2 of the career notes cover the longer version.

A wooden pier with a thatched palapa at sunset

Career & systems

  1. 2024–Present

    AI Lead Consultant

    Government of El Salvador

    AI strategy and clinical systems for DoctorSV, national telemedicine: triage, pre-consult summaries, chronic-disease monitoring. Began on USAID and EU-funded 5G coverage work.

  2. 2022–2024

    Staff Software Engineer

    Kavak

    Led the data engineering half of the pricing team after Kavak acquired OPI Analytics, owning the pipelines and cloud infrastructure behind vehicle pricing models in every market.

  3. 2018–2022

    Software Designer / AI Lead

    Globant

    Led the Mexico AI team across medical imaging, legal NLP, and anomaly detection. Consulting taught me to diagnose an organization’s real constraints before learning its politics.

  4. 2016–2018

    Data Science Manager

    Cultura Colectiva

    Built the data science and BI team from scratch. In September 2017 we paused normal work to build Verificado19s, a real-time civic verification platform for the Mexico City earthquake.

  5. 2016

    Data Science Lead · Data Lab

    Presidency of Mexico · Open Data Initiative

    Centralized technical team inside the federal government: find the bottleneck, build the analytical pipeline, and ship cloud infrastructure across multiple ministries.

  6. 2015–2016

    Data Scientist / Data Engineer

    Globant · OPI Analytics

    Discrete math optimization for Globant textile distribution: physical constraints, real money, no neural nets. At OPI, full-stack startup engineering and cloud ops from zero.

  7. 2010–2015

    Data Analyst

    Center for Emergency Attention and Citizen Protection (C4), Mexico City

    Crime prediction and officer evaluations in Excel macros and Matlab. The math was easy; the job was a dispatch number a commander off an 18-hour shift could act on at 6:00 AM.

Teaching

ITAM, 2020–2024

  • NoSQL for Data Science (M.Sc.)
  • Introduction to Programming for Data Science (M.Sc.)
  • Applied Analysis in Computational Mathematics

Selected publications

  • 2019Esquinca, Escalante et al. Recomendaciones para Sociedad Civil. Google Play Store.
  • 2016Martinez, Escalante, Beguerisse-Díaz et al. Understanding Human Behavior in Urban Spaces. IJWSR.
  • 2013Prieto, Escalante et al. Análisis Criminal y Uso de Cámaras de Seguridad en la CDMX. Buenas Prácticas para el Análisis Delictual en América Latina.

San Salvador now, Mexico City before that. I also run a specialty coffee shop (Aluxe) and build its operating software (CNTRL). If you are working on a problem where the model works on paper but the organization or operation is stuck, reach me by email.