# ChatDoc

> An AI tool that lets a public audit court's staff run semantic searches across rulings and regulations in natural language. Presented at AI.Brasil Experience.

- Organisation: TCM-GO
- Role: Mid-level Software Engineer
- Duration: Oct 2024 - Jun 2026
- Tools & Technologies: Python, NestJS, Machine Learning, NLP
- URL: https://adrianoleao.com/case-study/chatdoc
- Author: Adriano Leão (https://adrianoleao.com/)

## Context

The Tribunal de Contas dos Municípios de Goiás (TCM-GO) holds a large body of decisions, rulings and internal regulations. Finding the right passage with keyword search is slow when you don't know the exact words the document uses.

    - **AI.Brasil**: Experience, where ChatDoc was presented

    - **GAIA**: Google DeepMind's Brazilian Portuguese language model, explored with ChatDoc

## My role

Software engineer on a team working in partnership with CEIA/UFG. Depending on what the project demanded, I worked across frontend, backend, architecture, machine learning and data science.

## What I built

- A natural-language semantic search over the tribunal's decisions, rulings and internal regulations
- The full stack around it, from the interface to the backend services
- The machine learning and NLP components that power the search

## Outcome

- ChatDoc shipped to the institution's staff
- Presented publicly at AI.Brasil Experience
- Explored together with GAIA, Google DeepMind's Brazilian Portuguese language model

[Back to the profile](https://adrianoleao.com/index.md)