I've been writing code for years — automation scripts, web scrapers, data analysis, and everything in between.
A few of the things I've built on my own time:
A CLI and typed Python library that extracts a video's transcript, its comments (with likes, dates and replies) and its related-videos sidebar, writing clean, LLM-ready Markdown.
It drives headless Chrome through Selenium. Two optional modes ship with it: vl ask runs retrieval over what you already collected without re-scraping, and vl live analyzes a livestream chat in real time.
A CLI that tracks Claude plan consumption over time: it reads the official usage endpoint, stores snapshots in DuckDB, and reports burn rate, projections and 5-hour-window alerts.
It resolves the token through a fallback chain — an environment variable, the OS secure vault (Keychain, Credential Manager, Secret Service), or Claude Code's own credential — and refreshes OAuth by itself when it expires, never writing anything in cleartext.
REST endpoint that returns structured data for any company listed on B3 (Brazil's stock exchange) — pass the ticker via a ticker request header.
Built with TypeScript and Next.js, deployed on Vercel.
Open-source bot that uses Selenium to download company filings from Status Invest, summarizes them with the OpenAI API, and emails the summary.
Reading notes on Fluent Python, by Luciano Ramalho. I worked through Python Fluente, the Brazilian edition — so the notes themselves are written in Portuguese.
Python scripts and notebooks with my observations on the book's topics. I started from the first edition and kept folding in what I read in the second.
Terminal chatbot built on the OpenAI API.
Supports custom aliases and clipboard access.
Library for web scraping that exposes the results through an API.
Built with Selenium for browser automation and FastAPI for the HTTP layer.
Front-end app hosted on Vercel that calculates compound returns on recurring contributions. Built with React and Next.js.
Uses a Gmail app password and Python's smtplib to send email to any recipient.
Handy for having other bots and scripts deliver results to my inbox — company filing summaries, chatbot replies, or lecture notes.
Running on a Raspberry Pi, it can also send pre-parsed digests from news aggregators.
Turns source material into a multiple-choice quiz by constraining the LLM output to a structured quiz schema.
Uses NumPy to model an investment portfolio as a polynomial in (1+i) and solves for the return rate.