Success Story: Sales and Cattle AI JBS

MadeinWeb, as an AWS partner, assisted JBS with cloud processing and resource optimization.

JBS is a Brazilian multinational recognized as one of the global leaders in the food industry. Faced with high demand for products and the need to control sales, stocks and maintain quality in the 15 countries where it operates, the multinational sought solutions to ensure that its more than 230,000 employees could work more assertively.

Thus, in January 2020, JBS proposed a challenge to Made: to develop two internal web systems with the aim of having better control over sales, Sales AI JBS, and the company’s cattle, Cattle AI JBS.

The problem was that, before the project reached MadeinWeb, the multinational had already hired a company. This company was unable to meet the high demand that JBS needed, extending deadlines and presenting a series of structural flaws, which made the process difficult and did not allow its usability to be used to its fullest.

Among the main structural problems of the previous system, the programming code was the most worrying, since, due to deep flaws, the pages were taking more than 30 seconds to load, which made the system unusable.

To correct these problems, MadeinWeb divided the solution into two phases. The first aimed to correct usability flaws, stabilizing the platform.

CodeBuild was used together with CodePipeline to automate build and deploy processes to avoid failures, aiming for operational excellence during deployments. In this way, CodeBuild was responsible for integrating the data in Amazon S3, so that all of it was properly stored and approved.

It is worth mentioning that, in addition to these functions, CodeBuild was also responsible for integrating SAM and CloudFormation in the deployment of the user authentication API.

SSL/TLS certificates were then created and managed using AWS Certificate Manager, and all important credentials were securely secured using AWS Secrets Manager.

The second phase aimed to reduce loading time to ensure even more high-performance systems for users.

The first thing the team identified was that each Lambda had six to eight responsibilities within it, which hindered the performance of the systems and, consequently, increased loading time.

Thus, MadeinWeb divided the structure into more than 30 different lambdas, each configured with a different size and, consequently, cost, according to its function, optimizing the organization of the code and facilitating its maintenance.

To do this, the user authentication process was optimized using the resources provided by AWS Cognito. The website cache was also optimized and stored through Cloudfront, while the API cache, i.e. the Lambdas, was stored in Elasticache.

All infrastructure and codes were defined with Cloudformation, aiming at the automation and organization of resources and expenses.

For data consultation, the company used 100% of the Athena service, but over time, this solution was no longer viable for the institution’s needs at the time. The cost-benefit for JBS was no longer effective, and a change was necessary.

In this case, to optimize performance and costs, Athena was no longer used for queries. Instead, all data was imported into RDS (Relational Database Service), which is a high-performance relational database service!

To help optimize the system, we consume data produced through artificial intelligence and machine learning algorithms using DynamoDB.