I develop agent workflows, tools and integrations that connect LLMs to databases, APIs and business processes. Working with agents and AI automation is part of my day-to-day engineering, with a broader focus on technology for finance and wealth management.
My background combines social sciences, statistics, data consulting and BI development in the oil industry. I bring that analytical perspective to applied AI: turning business questions into useful software, structured data and repeatable workflows.
Beyond the public tools below, my project work includes:
- Agent orchestration: LangGraph and LangChain workflows, tool calling, asynchronous jobs and human review gates.
- Agent runtimes: persisted sessions, event-driven resumption, context compaction, bounded memory and token budgets.
- LLM reliability: typed outputs with Pydantic AI, source-grounded extraction and validation of generated entities and citations.
- Observability and delivery: structured execution traces, token and cost accounting, automated tests and containerized applications.
- Griphook — Python CLI for AI agents to query Microsoft SQL Server. Includes SQL AST validation to block writes, query limits, timeouts, schema exploration and structured JSON output. Ships with automated tests, a Claude Code skill and a Windows installer workflow.
- n8n-cli — REST API tooling and a Claude Code skill for inspecting and updating n8n workflows, investigating executions and retrying failures. JSON output and an optional read-only mode support controlled agent interactions.
- bert-finance-sentiment — Financial NLP project combining document collection, domain adaptation of BERT and LoRA fine-tuning for sentiment classification, with evaluation against baseline models.
- puc-engenharia-dados — Data engineering project using PySpark, SQL, Databricks and Delta Lake, with exports to AWS S3 and an analytics environment built with PostgreSQL, Docker and Metabase.
- ntnb-ibov-backtest — Python analysis comparing historical Brazilian equity and inflation-linked bond returns.
- Agents and automation: LangGraph, LangChain, Pydantic AI, tool calling, context management, n8n and Claude Code skills.
- Backend and data: Python, SQL, FastAPI, TypeScript, SQL Server, PostgreSQL, SQLAlchemy, Redis and asynchronous workers.
- Machine learning: financial NLP, BERT, Hugging Face Transformers, PyTorch and LoRA.
- Infrastructure and analytics: Docker, GitHub Actions, Databricks, PySpark, AWS S3 and Metabase.