Welcome to My Blog

Modeling the Relative Risk of COVID-19 Across the U.S.

Analytics
Spatial
We extend the Besag–York–Mollié (BYM2) model by combining its spatially structured and unstructured random effects with a dynamic temporal process within a Bayesian…
Feb 28, 2026
Jonathan Dekermanjian

How Machines Comprehend Language
Recurrent Neural Networks - Embeddings from Language Models (ELMo)

Natural Language Processing
We go over contextual embeddings and recurrent neural networks. Describing their inner workings and walking through an implementation of Embeddings from Language Models…
Jan 31, 2026
Jonathan Dekermanjian

How Machines Comprehend Language
Global Vectors for Word Representation

Natural Language Processing
A deep dive on the Global Vectors for Word Representation (GloVe) algorithm with an implementation in JAX.
Sep 13, 2025
Jonathan Dekermanjian
 

How Machines Comprehend Language
Skip-Gram with Negative Sampling Algorithm

Natural Language Processing
We explore the Skip-Gram with Negative Sampling algorithm and implement it in JAX from scratch
Aug 31, 2025
Jonathan Dekermanjian

How Machines Comprehend Language
Bonus: Text-Classification

Natural Language Processing
We expand on the hands-on example of the last post and show how we can leverage TF-IDF sparse embeddings to build a multinomial logistic regression classifier.
Jun 28, 2025
Jonathan Dekermanjian

How Machines Comprehend Language
Introduction to Vector Embeddings

Natural Language Processing
In this first post of a series on Natural Language Processing (NLP), we introduce the foundational techniques for converting textual language into numerical representations…
Jun 21, 2025
Jonathan Dekermanjian

Reproducibility & Scalability Part 4
Scaling With Ray

Engineering
In the final part of this series, we focus our attention on scaling the model training task. We begin by highlight the steps necessary to configure a Ray cluster. Followed…
Apr 17, 2025
Jonathan Dekermanjian

Reproducibility & Scalability Part 3
MLFlow Integration

Engineering
In part three of this series, we configure MLFlow for production by defining our tracking server as a containerized PostgreSQL database and our artifact store as…
Apr 12, 2025
Jonathan Dekermanjian

Reproducibility & Scalability Part 2
PyMC ModelBuilder

Engineering
Analytics
In part two of this series, we go over the necessary steps to take a time-series model into production by integrating the PyMC ModelBuilder class with Kedro.
Apr 2, 2025
Jonathan Dekermanjian

Reproducibility & Scalability Part 1
The Kedro Framework

Engineering
In part one of a four-part series, we are going to set up and walkthrough the Kedro framework. Kedro is an open-source Python framework that provides you with scaffolding to…
Mar 26, 2025
Jonathan Dekermanjian
No matching items