Projects

Efficient classification with vector embeddings

Efficient classification with vector embeddings 2025

This project enables automatic text classification using Large Language Models (LLMs), eliminating the need for traditional model-building approaches. It classifies any given text into predefined categories, making it particularly useful for organizing house renovation or rehabilitation expenses.

tagging classification vector-database aws machine-learning ai aws-lambda aws-cloudformation aws-sam huggingface pinecone
Read More
Hybrid search in Postgres

Hybrid search in Postgres 2024

Build a hybrid search system (semantic + keyword search) in postgres that prioritizes recency, making it ideal for dynamic domains like news and social media. It leverages PostgreSQL's pgvector extension to enable efficient vector search within the database, ensuring a balance between relevance and freshness in retrieval-augmented generation (RAG) applications.

retrieval-augmented-generation vector-database postgresql machine-learning pgvector docker hybrid-search add-recency-in-hybrid-search
Read More

Writings

School Vouchers Bot

February 20, 2025

Introduction The goal of this project was to integrate a school vouchers chatbot into a special report, marking our first public-facing, internally developed chatbot tool. To maintain journalistic integrity and prevent misinformation, we restricted the chatbot’s access to a curated...

RAG Weaviate Cohere OpenAI
Read More

Helpbot at The Texas Tribune

September 24, 2024

Problem Statement We use Slack at The Texas Tribune and like many other orgs out there we use a dedicated “help” channel for staff to get assistance on any IT problems, account access needs, office supply requests etc. This channel...

Engineering RAG AWS Bedrock Slack
Read More