Raul – Python, AWS, LLM

The project is a customized large language model (LLM) signed specifically for B2B clients. Unlike generic LLMs trained on vast, publicly available internet data, their model is tailored to each client's unique needs by being trained exclusively on their knowledge base.

• Oversaw the ingestion pipeline and managed the entire ETL process for storing data in Parquet format to be used by an LLM;
• Managed infrastructure using AWS and Terraform;
• Deployed and maintained container orchestration with Docker Swarm on EC2 instances;
• Automated deployment pipelines using GitHub Actions;
• Developed full-stack applications, focusing 80% on the backend and 20% on frontend development.

This project focused on collecting and consolidating user behavior to provide actionable insights for the marketing team.

• Built and maintained scalable cross-platform ETL pipelines, Integrating data from diverse sources;
• Designed efficient data processing workflows, reducing latency and improving performance;
• Automated data validation and error handling to ensure accuracy and consistency;
• Optimized queries and scripts for enhanced pipeline scalability;
• Deployed and managed ETL pipelines on AWS EC2 for cost-effective, reliable operations;
• Collaborated with cross-functional teams to deliver data solutions aligned with business needs;
• Enabled data-driven decision-making by providing clean, timely data for analytics and marketing;
• Contributed to Pinterest's growth by ensuring robust, scalable, and efficient data infrastructure.

Wallet integration between the bank's mobile app and third-party services.

• Worked on integrating third-party services into the bank's mobile application using REST APIs;
• Built serverless mobile applications using Node.js;
• Utilized AWS Lambda, API Gateway, CloudWatch, SNS, and SQS to develop and deploy scalable and secure integration solutions Developed and managed REST APIs to expose the bank's services to external parties;
• Utilized DynamoDB to store and retrieve data related to the integration;
• Collaborated with cross-company teams to ensure the timely delivery of integration projects.

The project involved developing and implementing a web-crawler to monitor and extract legal processes published in Brazil’s National Official Diary for a national bank client. The project aimed to streamline the identification, categorization, and routing of legal processes to the appropriate operational teams within the bank.

• Developed a web-crawler to extract juridical processes from Brazil’s National Official Diary;
• Implemented logic to filter and capture juridical processes related to a national bank;
• Applied machine learning concepts to analyze extracted data and identify common patterns;
• Automated the classification and categorization of juridical processes;
• Designed a system to redirect legal cases to the appropriate operational teams within the bank;
• Ensured accuracy and efficiency in processing legal documents through data-driven analysis.