Hello, i'm Marcos!
I build production-grade AI systems, specializing in LLM applications, retrieval-augmented generation (RAG), and agentic workflows.
My professional experience spans healthcare, biotechnology, life sciences, education, media, and financial services, with a strong focus on AWS infrastructure, reliability, and software engineering.
Currently, I lead the development of an agentic AI chatbot for a US biotechnology M&A platform. I've also worked on large-scale RAG systems processing millions of scientific and pharmaceutical documents.
LLM & Agentic Systems
- Production RAG architectures, retrieval pipelines, grounding, hybrid search, and reranking.
- Agentic applications, tool integration, MCP servers, and LLM orchestration.
- Evaluation pipelines, observability, latency optimization, and operational reliability.
Cloud & Data Infrastructure
- AWS: Bedrock, AgentCore, Lambda, Redshift, CloudWatch, S3, and EC2.
- Snowflake, Cortex AI, dbt, Databricks, and data engineering.
- CI/CD, deployment automation, and production monitoring.
AI-Assisted Software Engineering
- Coding-agent workflows with reusable skills, extensions, and guardrails.
- Automated quality checks, linters, commit hooks, and GitHub Actions.
- An emphasis on simplicity, maintainability, reproducibility, and developer productivity.
| Area | Technologies |
|---|---|
| Languages | Python, TypeScript, SQL, Bash |
| LLM engineering | RAG, MCP, LangChain, LlamaIndex, LangGraph |
| AWS | Bedrock, AgentCore, Lambda, Redshift, CloudWatch, S3 |
| Data | Snowflake Cortex AI, dbt, Databricks, PySpark |
| Software engineering | Git, GitHub Actions, Docker, CI/CD, REST APIs |
| Machine learning | Scikit-learn, PyTorch, Hugging Face, MLflow |
My personal configuration and development environment for the Pi coding agent, including custom extensions, reusable skills, MCP integrations, permission controls, and development workflows. Its my current coding agent and i love it!
My first extension for Pi agent. This one allows you to use your Anthropic subscription quota inside Pi, with full Terms of Service alignment.
A local AI-powered meeting analysis application, designed for private processing.
With so many blogs and posts to read it becomes difficult to not feel behind. This is my initiative to tackle this problem. My strategy: fetch, store, organize, cross-reference and summarize my "reading list" weekly.
Most of my production AI engineering work is performed in private enterprise environments. The repositories above reflect my personal experiments, tools, and engineering interests.
B.Sc. in Computer Science, São Paulo State University (UNESP), 2014.
Earlier in my career, I worked extensively with traditional machine learning, predictive modeling, recommendation systems, and data engineering. That experience continues to inform how I design and evaluate modern AI applications.
- Other Areas of Interest: Psychometrics, History, Tolkien's Middle Earth and Dragon Ball Z 😎 .
[