João Vitor Chaves de Souza

Building V4H, Wisecare's telehealth platform used by SUS, Brazil's public health system. I work on the Node and TypeScript backend (Express, TypeORM, dependency injection with InversifyJS, queues with BullMQ and RabbitMQ, real time with Socket.IO) and on the React frontend with Redux-Saga, plus the integration with CNES, the national registry of health facilities, and authentication through Keycloak.

I lead the computer vision applied to sports track at the Technology and Artificial Intelligence League, the first academic AI league in Paraíba: defining the projects in that area, following them technically and training new members. Part of the league since December 2025, first as a trainee.

Research, development and innovation with industry partners: ZOOX Smart Data (2025) and the Rômulo Passos education platform (current). Focused on LLM-based agents, computer vision, OCR and business intelligence.

Data engineering system that ingests heterogeneous reports from sports betting operators through a configurable pipeline: JSON metadata, vectorized cleaning and anti-fraud validation, loading everything into a PostgreSQL data warehouse. The flow is orchestrated by an event-driven architecture on Kafka, with a FastAPI gateway validated by JSON and compliant with the Brazilian data protection law.

AI platform that cut the production of annotated exams from one or two days to roughly ten minutes. The system extracts data from PDFs automatically and uses RAG agents to write commentary grounded in curated material, with an interactive chat and expert review (human in the loop) feeding continuous improvement.

Research in applied artificial intelligence, contributing to the group's projects at the Informatics Center of UFPB.

I led AI engineering at LembreMe, a conversational assistant that brings reminders, personal finance, goals, tasks and notes together in natural language, with calendar integration, in-house scraping for travel search, contextual recommendations and automated phone calls. The core uses LLMs for intent routing and response generation. In the same ecosystem, the meetings module transcribes in real time through Deepgram and extracts tasks, decisions, insights and open questions, delivering notes and an executive summary as soon as the call ends.

Agent generator that solved a bottleneck in CrewAI: instead of writing code, the user defines the agent's role, goal and context through an interactive interface, and the application generates production-ready code and configuration.

Intensive in-person program in Computer Science, with workshops and mentoring in software development.

First place among six teams in the program hackathon, with PixelMind.

LembreMe

Personal assistant that runs entirely on WhatsApp: finances, tasks, reminders, goals, and routines in natural language. Lead author, architecture and development.

Python · FastAPI · Node.js · whatsapp-web.js · LangChain · GPT-4o-mini

Pauta

Municipal political transparency: citizens report an urban problem, track its resolution, and get closer to the people who represent them.

Python 3.12 · FastAPI · SQLAlchemy · Celery · Next.js 16 · React 19

PixelMind

Video editor that accepts natural language commands, first place at the Trilha hackathon. Backend architect.

Next.js · TypeScript · Tailwind CSS · Node.js · Python · FastAPI

Vero

A clinical platform that follows the consultation in real time and runs the patient's pre-visit intake on WhatsApp. Lead author, clinical LLM pipeline, pre-visit intake agent, and document import.

TypeScript · React 19 · Vite · Tailwind CSS 4 · TanStack Query · Fastify