Engineering Lead · AI Product Engineering
I ship AI features to production, end to end: models, APIs, and the products people use.
Marcello Oliveira is a Brazilian software engineer and Engineering Team Lead at Deel, where he leads AI product engineering and ships LLM features to production, from the model call to the interface. He previously built software at Amazon, Coinbase and XP Inc., and is a three-time ACM ICPC South America finalist.
Stories
All 14 stories ›In Production · 7 minPrompt injection is our generation’s XSSThere is no textContent for a model: what XSS and XSRF teach about prompt injection, and why the defense moves from input to privilege.Read the story ›In Production · 7 minWe’re bad intuitive statisticians. So are LLMs.Monty Hall, a hostel in Foz do Iguaçu, and what temperature and top-p really do to an LLM’s next token.Read the story ›In Production · 8 min0.1 + 0.2 and someone’s salaryWhy money is never a float: IEEE 754 bits, payroll drift and rate rounding, shown one payslip at a time.Read the story ›
Work
Taking LLM features from demo to production at DeelWhat it takes to put a language model behind a button that people who run global hiring, payroll and HR will press. Patterns, not internals.Read the case ›Soma, XP’s design systemMulti-theme, technology-agnostic and multi-platform. I was Soma’s technical lead in 2020, and it launched publicly in a live session in February 2021.Read the case ›Arsenal: a micro-framework for micro-frontendsBuilt at XP Inc. on top of Webpack Module Federation, so many teams could ship parts of the same product independently. Deliveries got about 40% faster.Read the case ›Chameleon: CSS-in-JS for web componentsA theming engine for Stencil and web components, about 8x faster than styled-components in our internal runtime-theming benchmarks.Read the case ›Now
Updated November 2026
I lead the AI product engineering team at Deel. This is how we build with language models today.
What I am doing now ›- LLM features in production. We ship features that validate and generate job scopes and map job descriptions, inside products that people use to hire and pay teams around the world.
- A rules engine as the safety net. The model proposes; a deterministic rules engine decides what can go through. When the model is wrong, the rules catch it before a person ever sees it.
- Structured output, validated. Every model answer comes back as structured output and is validated against a schema before any code trusts it. Free text never flows straight into a product.
- One internal AI service. We are moving model calls to Deel’s internal AI service, with batching and rate limiting built in, so every feature gets the same limits, logging and controls.