
Forte Pulso
Free on AndroidWorkouts, food and health in one app, with your friends along.
5+ years taking data to production: ETL, databases and data lakes first, then ML models, and now LLM agents. I build the parts that make AI reliable: promotion gates, observability, guardrails and deploys you can roll back.
Multi-agent orchestration, tools over MCP and guardrails that keep agents on track.
Traceable training, promotion gated on quality and fairness, drift and zero-downtime serving.
Traces, tokens, cost and latency per model, on OpenTelemetry.
ETL, databases and data lakes: the foundation every model needs.
Some demos run in your browser; the ones that use an LLM need a login so I can keep usage in check.
Ask about the economy, the news, companies or Pix rules and watch the coordinator delegate to five specialists. Turn MCPs on and off and follow the full trace.
Ask a question →Write your own agent: instructions, model and MCP tools. Run it, save versions and compare two side by side, with every step and the trace.
agent-orchestratorCall the MCP server tools directly: CEP, CNPJ, holidays, Selic, IPCA and the dollar rate.
mcp-server-brasilAn agent builds the profile and you watch the MLflow champion's pipeline run step by step, or validate and score a CSV in batch.
mlops-platformUpload a spreadsheet, clean the data, compare models and export the Python code. Python and an AI assistant run in your browser; the file never leaves the device.
Opens at ml.vgmelo.tech ↗Multi-agent API: a coordinator delegates to specialists in parallel, with MCP tools, guardrails and end-to-end tracing.
MCP server with Brazilian public data for agents: CEP, CNPJ (alphanumeric too), holidays, business days, Banco Central and IBGE.
Self-hosted LLM observability: traces, tokens, cost and latency per model, with an OpenTelemetry GenAI endpoint and PII redaction.
End-to-end MLflow for credit risk: traceable training, champion/challenger promotion gated on quality and fairness, drift and serving that swaps models without downtime.

Workouts, food and health in one app, with your friends along.
Study music theory, rhythm and solfège at home, a little every day.
Brazil's company ID now accepts letters. How the check digit works with ord(c) − 48, what breaks in your systems, and a Python validator.
How I set up credit model promotion with quality, calibration and fairness gates, MLflow aliases and one-command rollback, plus the type bug that marked the champion as rejected.
A reasoning model returned empty answers and my orchestrator counted them as success. The two causes, the fix (one extra request, then an explicit error) and the tests that catch it.
I consult on agents and MCP, MLOps, LLM observability, data, and app development.