Dashboard // Welcome
// Hello, I'm

Welcome to my Dashboard

Data & AI leader who still ships code. I own a company-wide data platform end to end — warehouse, lakehouse, orchestration, governance — and build the production LLM systems that run on top of it: agent-safety layers, natural-language interfaces to the warehouse, and customer-facing conversational agents.

Profile // About

I work across the whole stack of a data organisation: a nine-schema Redshift warehouse fed by 16 source systems, a 99-table Athena/Iceberg lakehouse, the orchestration and governance layers around both, and the applied-optimization and LLM systems built on top — from Bayesian marketing mix models and convex budget allocation to agent-safety libraries and text-to-SQL agents.

My career spans Brazilian agencies (WMcCann, IPG Mediabrands), media (Rede Globo), an AI startup (MINT), and US e-commerce (Filterbuy), alongside an independent data & AI practice. Engineering by training: MBA in Data Science (USP), MSc in Industrial Engineering (COPPE-UFRJ, interrupted), dual BSc in Mechanical Engineering and Physics (UFF).

Email
rodrigo@rodtex.dev
Website
rodtex.dev
Languages
PT (Native), EN (Fluent), ES (Intermediate)
Location
Sao Paulo, Brazil
Metrics // Skills
Data Engineering
SQL & Dimensional Modelling 95%
Redshift / BigQuery / Athena 92%
Apache Iceberg / Lakehouse 85%
Data Governance & Quality 88%
AI & LLM Engineering
Agent Tool / Skill Layers 92%
Production LLM Agents 90%
Text-to-SQL / NL Interfaces 88%
MCP & LLM Safety 85%
Cloud
GCP 90%
AWS 85%
Docker 70%
Applied Optimization & ML
Convex Optimization (cvxpy) 88%
Bayesian Modelling (Meridian) 88%
Forecasting & Predictive ML 82%
Orchestration & DevOps
Event-Driven Pipelines 92%
Matillion / Step Functions 88%
CI/CD (GitHub Actions, OIDC) 88%
Terraform & Docker 80%
Marketing Science
Media Mix Modeling 88%
Attribution Modeling 85%
Audience Segmentation 85%
A/B Testing 82%
Timeline // Experience
2024 — Present

Marketing Analytics Director

Filterbuy · Fort Lauderdale, FL (Remote)

Own the company-wide data platform for a D2C and marketplace e-commerce business: a nine-schema Redshift warehouse fed by 16 source systems, a 99-table Athena/Iceberg lakehouse, and the governance and quality layers around both. Built the Amazon advertising decision platform end to end, a 61-skill agent-safety library for LLM agents, and a Bayesian marketing mix model. Lead engineers and analysts cross-functionally.

2025 — Present

Principal Consultant — Data & AI

rt analytica · Sao Paulo, Brazil

Independent data and AI practice serving e-commerce, food-service and retail clients. Built a multi-tenant LLM agent platform on Cloud Run, an event-driven ERP-to-warehouse pipeline handling a five-second vendor SLA, a natural-language analytics agent over BigQuery, and an open-source Model Context Protocol server.

2022 — 2023

Head of Data LATAM

IPG Mediabrands · Sao Paulo, Brazil

Led data strategy and analytics operations across Latin America for one of the world's largest media agency networks. Oversaw cross-functional teams delivering data-driven media planning and audience insights.

2020 — 2021

Data Manager / Tech Lead

MINT · Sao Paulo, Brazil

Built AIRE, a next-best-investment engine using Markowitz-style portfolio optimization to allocate media budget across channels and DSPs, increasing ROI by 25%+. Also built Megalista, privacy-safe ML-driven audience generation, plus multi-touch attribution models.

2019

Data Analytics Manager

WMcCann · Sao Paulo, Brazil

Led a marketing mix modelling study for General Motors ($100M+ annual marketing spend), defining optimal annual budget allocation, with automated measurement pipelines in SQL and BigQuery.

2018

Data Scientist

Rede Globo · Rio de Janeiro, Brazil

Automated the reporting infrastructure for Gshow, Latin America's largest entertainment portal, reducing manual reporting time by 80%. Built a churn optimization algorithm for content engagement.

2017

Data Science Consultant

Bridge · Rio de Janeiro, Brazil

Delivered data science consulting projects spanning statistical modeling, process optimization, and analytical tool development for diverse industry clients.

Records // Education
MBA in Data Science & Analytics
USP — Universidade de Sao Paulo
2020 – 2021
MSc in Industrial Engineering (interrupted)
COPPE-UFRJ — Federal University of Rio de Janeiro
2018 – 2019
BSc in Mechanical Engineering
UFF — Universidade Federal Fluminense
2010 – 2015
BSc in Physics
UFF — Universidade Federal Fluminense
2006 – 2010
Projects // Portfolio

Amazon Bid Optimization Engine

Automated bid and budget optimization for Amazon Advertising using serverless AWS Lambda functions with real-time performance feedback loops.

Python AWS Lambda Amazon Ads API

Causal Impact Analysis

Incrementality testing frameworks using geo-holdouts and matched market experiments to measure true causal lift of marketing campaigns.

Python CausalImpact Geo-Experiments

Vertex AI Recommendations

Product recommendation engine on Google Vertex AI for e-commerce, leveraging collaborative filtering and BigQuery ML.

Python Vertex AI BigQuery

Marketing Mix Model

MMM using Google Meridian framework for media budget optimization with Bayesian inference and scenario simulation.

Python Meridian Bayesian

Data Warehouse

End-to-end data warehouse design, build, and maintenance on Redshift and BigQuery — ETL pipelines, schema optimization, and 40% query cost reduction.

Redshift BigQuery Matillion

Predictive LTV Model

24-month customer lifetime value prediction using RFM features and XGBoost, integrated with server-side pixel for media optimization.

Python XGBoost RFM

Google Ads Automation

Automated Google Ads campaign management scripts using the Google Ads API and GAQL for bulk operations and performance tuning.

Python Google Ads API GAQL

Server-Side GTM

Server-side Google Tag Manager implementation that uncovered 30%+ conversion undercounting from ITP/ad blockers, driving first-party data infrastructure.

GTM Cloud Run Meta CAPI

Agent Safety Skill Library

61 agent-callable skills across three ad platforms behind one mutation-safety contract — dry-run by default, risk gating, approval and audit.

TypeScript Zod LLM Tool Use

Multi-Tenant LLM Agent Platform

Production conversational agents on Cloud Run — customer-facing ordering over WhatsApp plus internal BI-analyst agents, each tenant with its own guardrails.

TypeScript Cloud Run MCP

Event-Driven ERP Pipeline

Rearchitected an ERP-to-warehouse integration that was failing a vendor's five-second webhook SLA at p50 19s into sub-second acknowledgement.

Python Pub/Sub BigQuery

Text-to-SQL Analytics Agent

Natural-language questions answered against a BigQuery warehouse, plus a guarded write path to a live ERP behind an action-safety layer.

Python FastAPI BigQuery

blingMCP — Open Source

An MIT-licensed Model Context Protocol server consolidating 254 ERP API endpoints into 18 LLM-friendly tools via OpenAPI code generation.

TypeScript MCP Open Source

LLM-Assisted ETL Diagnostics

Exported a GUI ETL platform to 361 greppable SQL artifacts, grounded an agent on a hand-built schema dictionary, and drove a deterministic diagnostic runbook.

Python Claude Redshift

Marketing Funnel Scoring

A weighted 0–1000 channel scoring engine with LLM-generated recommendations reviewed through a human approval application.

Python BigQuery Claude