PaulaVelosa

Rooted in Colombia, I build digital worlds from code, image and movement — experiences that carry memory, identity and imagination forward.

Work with me→See my work

↟ a living cosmology built from code, image and movement

structure: structure
transformation: transformation
movement: movement
memory: memory
craft: craft
encounter: encounter

descend

codestructureillustrationimage3Dspacedancemovementwritingmemory

Four ways of making

Every discipline is a different way of shaping culture.

“digital painting — after Basquiat”

craft

Made by hand

Personal work exploring bodies, symbols and imagined characters — made outside client constraints.

See everything→

Fresh carvings

Every project leaves a mark. New pieces added to an evolving archive.

See everything →

structure— Dec 2025

Avatar Flower

An AI logistics platform for Colombia's flower export chain — from farm to airport cold room. Agentic workflows, tablet-first UX for six operational roles, and offline-tolerant flows for freezing rooms.

structure— Jan 2024

Rentoso

Where I first put LLMs into production: tool-calling UIs that make a model's actions visible and trustworthy, plus the retrieval infrastructure behind them — embeddings, a vector database and a RAG pipeline.

Field notes

Memory in progress — where the other territories get written down.

See everything →

01 · Jun 2026

Designing for frozen hands: an interface for the cold rooms at El Dorado

The users I was designing for wear gloves, stand in near-zero temperatures, and work in a corner of the airport most travelers never see. What they taught me about interfaces fits in one sentence: the room always wins.

02 · Jun 2026

Speccing is 80% of the work (code is the easy part)

Roughly 80% of the time that produces a good feature is spent before and around the code — understanding, specifying, prototyping, revising. Code is increasingly the cheap part. Some of mine is AI-generated. The spec never is.

03 · May 2026

Thinking in state: what LangGraph taught me about agents

My first agentic workflows were duct tape — call the model, parse, retry, add a flag, add another flag. LangGraph forced a different discipline on me: think in state, not in prompts. The model is the easy part.

encounterLet's light a fireBring an idea. Let’s see what we can ignite — I read everything and reply within 24 hours. ↯