Juan Luna

I developed a Chrome extension that enables semantic search and AI analysis of web pages and PDFs directly within the browser. The extension captures documents from the user's browsing session and sends them to a Python back end I built, which chunks the content, generates embeddings, and stores them in a vector database.

The primary challenge was to create a seamless integration between the browser and the back end while maintaining fast response times. I implemented an efficient chunking strategy to preserve document context and enable accurate semantic search across large document collections. Users can search using natural language queries that understand meaning, not just keywords, with results enhanced by LLM features like summarization. Search results appear as highlights directly within the original documents, making it easy to find relevant information in context. I also added document-level AI operations, enabling users to generate summaries or create quizzes from entire files with a single click.

This project successfully streamlined information retrieval and eliminated manual document review through intelligent automation, significantly improving user productivity.

A full-stack, multi-tenant application using React/Python to manage Vertica databases. The application managed:

• Database installation and basic configuration
• Automatic scale up and down
• Automatic backups

The scale behavior was done using a mixture of Kubernetes HPA (horizontal pod autoscaling) and Python.

An ETL project for a big manufacturing company to daily process millions of rows, pulling them from an Oracle database, transform then and create data ready to be used in PowerBI reports. The task management is done in Dagster (using Python for the logic) and deployed in Kubernetes in Azure Kubernetes Service (AKS).

An Apache Airflow pipeline built for a gaming company to process millions of live data records coming from pc/tablet/phone applications. The process transformed the rows and provide live analytics to the company.

The pipeline was automatically deployed in AWS as an ECS service with GitLab CI and the base infrastructure was created with Terraform.

A Scala/Drools points calculator built to be used by Qantas back-end services to calculate egular trip earned points and consumed points and apply promotions.

The application replaced the current implementation and reduced the process time from minutes to seconds, increasing back-end responsiveness.

A full-stack application, built with Java and React, used to fill a Colombian law requirement of offering customers a fair opportunity of translating from a security pension provider to another.

The application integrated data from different pension provider companies into a single data source and calculated the pros and cons of changing providers.