Rafael works with Java Spring Boot to build, test, and deploy software solutions to the cloud to meet client needs. He uses his experience in system architecture and SDLC to deliver distributed, scalable, enterprise-grade, low-latency solutions using modern technologies. Rafael also migrates applications to the cloud and configures web containers on solutions. In recent years, he has been specializing in Machine Learning, NLP, and Generative AI using LLMs. His expertise includes advanced techniques in RAG, multi-agent systems, Agentic workflows, agent construction, and model fine-tuning. Using LangChain, LangGraph, LangSmith, CrewAI, and LlamaIndex, Rafael is passionate about building intelligent systems that solve complex problems by combining state-of-the-art LLMs with multi-agent architectures.
AI, ML & LLM
Machine Learning
Large Language Models (LLMs)
Generative AI
CrewAI
LlamaIndex
Langsmith
Langgraph
LangChain
Model Fine-tuning
Agentic AI
AWS Bedrock
Airflow
Identified and evaluated potential projects for AI application within portfolio companies, leading the initial phase of AI use case ideation and validation.
Developed the POC for the Document Tagging Automation project using vision models for field identification, LLMs for context extraction, and embedding vectors with advanced reranking techniques for tag mapping.
Successfully presented the POC, demonstrating the solution's viability and securing approval for full project development at CAIS.
Developing and implementing the AI automation component of the Document Tagging Automation project.
Scaling the solution to production based on the developed POC, achieving over 93% accuracy in mapping and reducing human validation time from 2-3 days to less than 2 hours per document.
Developed microservices and RESTful APIs using Spring Boot, Flask, and FastAPI.
Worked on cloud services with AWS EC2 and ECR and on database management with PostgreSQL and MongoDB.
Led a team of data scientists and engineers, spearheading a groundbreaking project focused on developing a text-to-SQL solution.
Built language models using ML and NLP and applied LLM models, fine-tuning techniques, and prompt engineering to optimize the performance of the NLP solution.
Used SageMaker, DynamoDB, Amazon Bedrock, and the Amazon Knowledge Base for seamless development and deployment.
Implemented Docker for efficient containerization, ensuring a streamlined and scalable workflow.
Drove the development of a cutting-edge solution that converts natural language queries into SQL queries, revolutionizing data interaction.
