With Charla, built by MadeinWeb on AWS, the port-industrial complex speeds up environmental licensing using RAG over an archive of thousands of pages.
Yohan Consani · Published
Porto do Açu, one of the largest port-industrial complexes in Latin America, has begun testing generative artificial intelligence to optimize its documentation processes. The solution, developed by MadeinWeb on AWS infrastructure, produced a generative chat called Charla that speeds up the environmental-licensing process.
One of the main challenges of an operation at this scale is handling a huge volume of documents. Finding answers within an archive where a single document can run to more than a thousand pages made searching slow, and it affected response times for regulatory queries. The EIA (Environmental Impact Study) documents gather data in varied formats such as .txt, .pdf, .xlsx, .docx, .tiff, .json, .jpeg and .png.
The solution
The solution is based on a Retrieval-Augmented Generation (RAG) system that searches a vast set of data provided by Porto do Açu and makes relevant information easy to access with precision. In practice, the Charla chat works like a virtual library of technical manuals, standards and EIA data, enabling fast, accurate queries without having to page through hundreds of pages by hand.
The project began with the choice of Amazon Web Services (AWS) and is led by a six-person team from the Licensing and Innovation areas, dedicated to evaluating the use of generative AI for process optimization. The combination of AWS Lambda and Amazon S3 let the solution handle a large volume of data and requests without compromising performance, while ensuring the integrity and security of the information.
Results
The implementation centralized and automated queries, cutting response time in licensing processes and increasing team productivity. By removing the need to search manually across multiple databases, the Charla chat delivered a faster query experience and freed teams up for quicker decisions.