Turning scattered web data into supporter insight
Python and FastAPI pipelines that read unstructured web pages and documents and turn them into structured insight about nonprofit supporters.
- Focus
- LLM pipelines
- Data extraction
- Stack
¿Prefieres español? Lee esta página en español
Hi, I'm Daniel, an AI engineer in Colombia. Pandora's Lab helps teams in the US and worldwide turn promising AI experiments into dependable software their customers can count on.
Consulting to find where AI pays off, or a build from first version to production. See the work
How I work
You see working results early, and nothing goes live until it's dependable. Each step adds switches to this little office. Turn them on and watch the runs settle onto one safe path.
One call: what a good result looks like, and what must never happen.
Small and working, on your real examples, with answers in a shape your systems can use.
Tests, checks on real cases, and limits on what it can touch, before anyone relies on it.
Monitoring flags quality drift before your customers notice.
Keys inside the sandbox: 1–6 flip a switch, B turns them all on, R resets.
No switches on yet. The agent is guessing its way through the office.
Starting point: 3/10
Before, with no switches
Off. Add Evals to score every run and discard the bad ones.
Off. Add Watch it live to see every run as it happens.
Drift caught: 0
Work
From city hall to the nonprofit sector: where, when, what I built and what I built it with.
Python and FastAPI pipelines that read unstructured web pages and documents and turn them into structured insight about nonprofit supporters.
Full-stack product features shipped end to end with product and design, for nonprofit fundraising tools.
Led the Citizen Relations Office's digital transformation and was Product Manager of the District's appointment-scheduling tool. Later, internal tools in React and Next.js.
Migrated the company website across React, Node, Hugo and Go.
Work together
Start with a plan or go straight to a build. Either way, we start with a 30-minute call.
A clear plan for where AI helps your team, where it doesn't, and what it takes to run it safely.
A product, pipeline, agent or internal tool built end to end, tested and monitored from day one.
About
I'm an AI engineer at GoodUnited, building LLM pipelines for the nonprofit sector. Through Pandora's Lab I take on consulting and build projects for teams in the US, Latin America and beyond.
I started in frontend, moved to full-stack, and now build AI systems. I care most about technology with a social impact.

Blog
Most AI agent projects get canceled over cost, unclear value or missing controls. When you actually need an agent, what to build first, and how to shape it so it's dependable.
An agent that's right 90% of the time fails far more than it seems. How to build evaluations for AI agents from real tasks, what to grade, and why to read the transcripts.
Tools are how an AI agent acts on your systems, and the part that changes its results the most. How to choose them, name them, describe them and make them hard to misuse.
FAQ
No. We start from your process and goals in plain language. I handle the technical side and explain the trade-offs as we go.
No. I do full-stack product and data work too, and use AI where it earns its place.
Anything from a focused review of one process to an end-to-end build, tested and monitored. We size it together on the first call.
We go through your processes together, find where AI saves real time or money and where it doesn't, and you leave with a concrete plan: what to build first, what data it needs, and how to run it safely.
Yes. I work remotely with teams in the US and worldwide, in English or Spanish. Colombia is on UTC−5 all year, so our working hours overlap with US Eastern and Central time.
A 30-minute discovery call. Email hello@pandoraslab.dev or message me on WhatsApp.
Contact
Tell me about the process that keeps surprising you. We'll start with a 30-minute call and find the first thing worth fixing.
It helps to mention:
hello@pandoraslab.dev