Software Engineer · Frontend, Mobile & AI
This site is white-label too — flip the design switcher in the header and the entire skin changes. Same layout, different product.
// 01 · projects
Local-first personal finance backend with an agent-native MCP layer. A chat that answers questions about your own data by calling your own tools, with a provider-agnostic LLM layer. Built AI-assisted in ~3 weeks; in daily use.
One React Native codebase, ten independently branded apps shipping to both stores with their own pipelines. 50K+ users monitoring IoT devices in real time.
export default function DatePicker(props: DatePickerProps) {
const state = useDatePickerLogic(props);
return <DatePickerView {...state} />;
}The design system the entire Ubidots platform runs on — ~25 components on Radix primitives, CSS-variable tokens, and a 3-layer component spec. Built solo; every new feature ships on it.
const { can } = useFlags();
if (!can("billing.invoice.approve")) {
return <Forbidden />;
}A custom authorization layer spanning three internal applications: ~10 roles, ~30 flags, own admin UI. Built in a week after evaluating off-the-shelf options.
const user = await findUser(id);
if (user.isOk()) render(user.unwrap());
const timeoutInMinutes = chrono(2).hours().toMinutes();Rust's Result and Option types brought to TypeScript, and a chainable time-conversion library running in production in an app with 50K+ users.
// 02 · experience
// 03 · skills
// 04 · about
I'm a software engineer based in Medellín, Colombia.
I started out building e-commerce for agency clients. Then I led mobile development for an IoT platform, where I designed a white-label architecture that turns a single React Native codebase into ten independently branded apps — now used by more than 50,000 people to monitor their devices in real time. Along the way I built the design system that the entire platform runs on today.
These days I work across the stack: React and Django at Ubidots, and React, Next.js, Node and FastAPI on internal products for a US strategy consultancy.
What I enjoy most is understanding the why behind technical decisions. I'm not satisfied with something just working; I want to know why it works, and make sure whoever comes next can understand it without struggling.
Lately my work has gone AI-native — the team restructured around it, and I've been building with LLMs myself: an MCP server and a provider-agnostic model layer.