Home Projects Portfolio Dashboard Export PDF Log in

Languages

English
65

Latest Updates

Documenting code, one commit at a time.

Python 10 posts
×

Scaling Medical AI: Building a Robust MLOps Pipeline for Diagnostic Analysis

Architectural Overview

In the health-mlops project, the focus has shifted toward building a high-performance, reproducible machine learning pipeline for medical image classification. Specifically, we have implemented a streamlined MLOps architecture capable of achieving a ROC-AUC score of 0.9937, ensuring highly reliable diagnostic predictions.

The MLOps Pipeline Pattern

0 GitHub Actions Python

Optimizing GitHub Actions: Moving from Containers to Direct Execution

Managing CI/CD pipelines often involves balancing build speed with environmental consistency. Recently, while working on the mlops-credito project, I encountered a bottleneck where our pipeline execution overhead was becoming a friction point for rapid testing. This is the story of how simplifying our runner configuration improved our feedback loop.

The Problem: The Container Overhead

0 Python

Establishing a Foundation for Predictive Modeling

In the ongoing development of the project prediccion-abandono, a significant milestone has been reached: the integration of the initial model configuration file. This project is dedicated to exploring predictive analytics, specifically focused on identifying patterns related to user churn.

Establishing a Robust Pipeline for Predictive Modeling

Introduction

The prediccion-abandono project is designed to tackle churn prediction, focusing on identifying patterns in customer behavior to inform proactive retention strategies. As we initialize this project, the primary objective is to lay a solid architectural foundation that supports data transformation and model lifecycle management using Python.

The Pipeline Pattern Approach

0 Python Pandas NumPy

Architecting Automated Schedules: Building the Hexapod Planner

Structuring Robotic Routines

Managing complex, multi-legged robotic systems requires consistent execution of movement patterns and behavioral tasks. The hexapod-planner project was initiated to provide a structured approach to defining and executing these regimens, ensuring that hardware operations follow a predictable, time-based schedule.

The Core Logic: Data-Driven Planning

0 Jupyter Python

Structuring Data Science Portfolios with Jupyter

The Goal

Organizing data science experiments and findings can quickly become a disorganized mess of files and stale results. In the data-science-portfolio project, I recently focused on establishing a more systematic approach to uploading and cataloging my research artifacts.

The Approach

To ensure that my analysis remains reproducible and accessible, I have adopted a clean

0 Jupyter SQLite Python

Structuring Data Science Workflows with Jupyter and SQLite

Introduction

In the data-science-portfolio project, the focus has been on organizing research assets and analytical experiments. Managing growing datasets within a portable and robust format is essential for any reproducible data science workflow. This post explores the approach of integrating Jupyter notebooks with SQLite to maintain clean, queryable project archives.

The Workflow

Modeling Solar Cell Structures: A Data-Driven Approach

Architectural Overview

In the perovskitas-para-celdas-solares project, we have been working on refining the computational modeling of solar cell structures. Accurate modeling is essential for predicting the efficiency of perovskite-based cells, and this requires a systematic approach to data transformation and analysis.

The Workflow

To move from raw simulation data to actionable

Streamlining Fraud Detection: Lessons from Pipeline Optimization

Improving Data Visibility

In our project pipeline-deteccion-fraudes, we recently revisited our monitoring and dashboarding strategy. When building a pipeline-driven fraud detection system, having "blind spots" in your metrics isn't just an inconvenience; it's a security risk. After iterating on our infrastructure, we have smoothed out our dashboarding layer to better reflect the underlying