vol. 01 — essay
The AI industry is selling one thing. Enterprises need another. The gap between the two is becoming the most important story in tech. On the kingdom in the clouds — and the town that pays its rent, by the token.
Mozilla can swap a new model in the day it ships because the orchestration logic, feedback loops, and project context are already there. The cleanest case study of why the orchestration layer is where the real work lives.
AI can handle much of routine junior work — but cutting the entry-level pipeline damages leadership development. The smart path is hybrid roles, not replaced roles. Written off an HBR piece on the same theme.
Two-model AI LinkedIn-content product, built end-to-end concurrent with the MSBA. Spark is a 6-agent system that drafts a polished post in ~30 seconds. Astra grounds posts in the user's professional context plus live news from NewsAPI. Agent design, orchestration, prompt engineering, and the user-facing interface — all me.
A real-time financial-events tool that ingests primary sources and produces structured economic-impact assessments. The interesting layer isn't the news capture — it's the cascade reasoning: when X happens, what else moves, and how confident is the model in each link of the chain.
Built for the SMU MSBA cohort: students ask plain-English questions and an LLM with retrieval-augmented generation translates them into SQL, executes against the alumni database, and returns matching profiles. End-to-end ownership of the data layer, retrieval, LLM orchestration, query generation, and the interface.
Built from scratch for an early-stage healthtech startup. An internal discovery and qualification tool that combines a rules-based scoring engine with an AI-enhanced grading layer (Claude API), enriched by two clinical-trial data APIs. Translated a CEO conversation into operational scoring logic. Remained in use after the engagement.
An 8-script SQL pipeline (ingestion → cleaning → features → risk scoring) over 16K accounts across two seasons. Churn model at 58.1% precision flagged accounts representing $500K+ in projected at-risk revenue. Four counter-intuitive renewal findings via hypothesis testing. Power BI executive dashboard. P1/P2/P3 intervention plan delivered to senior stakeholders.
Unit economics, scenario models, pricing strategy, NAICS / HS code classification, and FDA / USDA compliance memos — distributor vs. e-commerce vs. JV structure comparison for clients in consumer products, food & beverage, and B2B services. 5+ live client engagements over 12 months.
I sit at the seam where strategy meets software — and I work best forward-deployed: with the operator, inside their problem, until the thing is in production.
MS Business Analytics, SMU Cox (May 2026). BS International Economics, TCU (May 2024). Founder of Heureka. Trilingual — pt · en · es. Brazilian & Italian citizen with EU/EEA work authorisation.
SQL pipelines · financial modelling (unit economics, FX, scenario, break-even) · AI integrations (Anthropic Claude, multi-agent orchestration) · customer discovery, ICP scoring, competitive intelligence.
Brazilian national golf athlete; two years in the Brazilian Navy's Olympic Sports Program. A decade of high performance under structured pressure, carried into analytics and product work.