Intern

• Built institutional web systems with PHP, JavaScript, HTML, CSS, and MySQL; also handled support, maintenance, networking, and structured cabling.
• Configured Linux environments and containerized internal apps with Docker on a dedicated server.
• Built an agentic procurement pipeline (Brave API, scraping, Gemini 2.5 Flash, vector database) for bid-compatible products and suppliers.

TypeScript, Next.js, React, PostgreSQL, Prisma, NextAuth, Vercel

• Production web application for a personal training studio, with admin and student areas, scheduling, members, finance, workouts, anamnesis, and authentication.

Swift, Core Audio, C, SwiftUI, macOS

• Native menu-bar app for real-time in-ear mic monitoring with ultra-low software latency (Core Audio HAL, lock-free ring buffer).

TypeScript, Fastify, PostgreSQL, pgvector, AI APIs

• Procurement product-matching API that normalizes tender items, searches and extracts offers, scores evidence, checks suppliers, and judges candidates.

Python, PyTorch, LoRA, GRPO, Qwen

• Research project training a small coding agent to repair bugs through tool calls while reducing token and tool-output usage.

Swift, AVFoundation, AppKit, macOS

• Native always-on-top circular camera overlay for screen recordings and presentations, distributed as a signed DMG.

React, TypeScript, React Router, Vite

• University project: a local movie catalog with search, typed components, dynamic detail routes, shared layouts, and a custom not-found page.

Swift, EventKit, TypeScript, MCP

• Local MCP server exposing safe Calendar queries and mutations through macOS EventKit, including recurrence and timezone support.

Java, Swing, JPA, Maven

• University project: a desktop CRUD management system for clients, contacts, suppliers, and products, with persistence through JPA.

Swift, AppKit, Codex, Claude Code, macOS

• Native menu-bar app for monitoring Codex and Claude Code usage limits across accounts, distributed through versioned releases.

Python, NumPy, PyTorch

• Transformer implementations in NumPy and PyTorch covering attention, feed-forward layers, normalization, residuals, and character-level training.

Rust, CHIP-8, desktop graphics

• CHIP-8 emulator implementing ROM loading, instruction decoding, execution, display, keyboard input, and timers.

Python, MoviePy, uv

• Command-line script that converts a source video into a silent timelapse with a configurable target duration.

Software Engineering

DeepLearning.AI & Stanford · Coursera

Luiz Otávio Miranda · Udemy

• Languages: Native Portuguese · Full professional English · Intermediate Spanish
• Technologies: TypeScript, JavaScript, PHP, SQL, HTML, CSS, PostgreSQL, MySQL, Node.js
• Tools: Claude Code, Codex, Cursor, Git, GitHub, Vercel, Docker, Linux, Vitest
• Frameworks: Next.js, React, Tailwind CSS