Hello, and welcome.
I’m Jeison (it sounds like “Jason”), a statistician and an AI engineer at Globant, where my client is Disney and I work on the agent design of LLM systems. I work in R and Python, and I like data that moves through time.
I studied statistics at Universidad del Valle, and my thesis compared ways of estimating the fractional differencing parameter in long-memory time series. I then spent five years at CIAT, the International Center for Tropical Agriculture, on climate and agriculture: crop simulations, the models behind AClimate Colombia’s seasonal forecasts, a shared UNFCCC Momentum for Change award in 2017, and papers in Climate Services, Frontiers in Ecology and Evolution and Computers and Electronics in Agriculture. In 2020 and 2021 I taught time series, probability and statistics at Univalle and Universidad Católica. Since 2021 I have been at Globant, on churn, segmentation, pricing, forecasting and recommendation. The thread through it all is state-space models, Kalman filters and honest uncertainty.
I live in Cali, Colombia, and my time series curiosity has even reached music, through a thesis I co-directed on the spectral analysis of cumbia.
This site is built with Quarto and hosted on GitHub Pages, and the source is on GitHub. Posts are CC BY 4.0 unless stated otherwise. The figure next to my name is a simulation of a Kalman filter tracking a hidden state from noisy data.
More in my CV and Publications.