The data exists, nobody acts on it
Reports that get opened once a month and change nothing. The information is there; what's missing is the system that turns it into a specific action with a name attached.
Data & AI engineering
Data and AI engineering for businesses that need decisions, not dashboards.
I'm a financial engineer. I've spent a decade in risk and data, and I build the software and AI systems that turn that experience into decisions you can act on.
The problem
You have data. What you don't have is decisions.
Manual work that doesn't scale
Reconciliations in Excel, invoices someone retypes by hand, figures copied from one system into another. It works right up until the volume grows.
Tools that don't talk to each other
The CRM doesn't know what the ERP knows. Every integration you haven't built is a person doing by hand what a system should be handling on its own.
Services
Three ways to work together.
01 — Highest value
AI & automation systems
RAG systems over internal documents, LLM-powered agents, workflow automation and document processing pipelines.
Not a chatbot bolted onto your website. A system that reads your documents, understands your context and answers with the source cited — or one that runs the whole process end to end without anyone opening a file.
Let's talk about your AI system
02
Custom backends & data applications
APIs, payment gateway integrations (Wompi, Stripe), booking systems, internal dashboards, ETL pipelines and forecasting models.
This is where the financial engineering shows. A forecasting model that understands seasonality and cash flow is not the same as a generic one, and a reconciliation built properly knows what it means when two figures don't match.
Tell me about your project
03
Conversion-focused websites
Landing pages and small business sites built to generate qualified enquiries.
Fast, measurable and built around one goal. No bloated templates, no pages that take five seconds to load.
Get your landing pagePortfolio
What I've built.
Lab
Systems I built to demonstrate what's solvable. Each one is functional and you can try it.
CreditSentinel
- Problem
- Getting the decline right isn't enough. Regulated lending has to explain why, with a reason that holds up to a regulator. Most systems hand over a score and leave the explanation to a person.
- Approach
- A service that scores probability of default, turns it into approve, decline or review, and issues the reason code grounded in the model's own SHAP drivers plus retrieved lending policy — never improvised. Around it: a versioned model registry, a model card, calibration and fairness metrics, and PSI drift monitoring.
- Data
- Synthetic, generated from a pinned seed. No real people and no personal data.
TerraSignal
- Problem
- Forecasting a renewal rent on a commercial unit and estimating whether the tenant will default are two expensive decisions usually made with a spreadsheet and judgement.
- Approach
- A platform that models both under one explicit rule: language models interpret, they never compute the financial maths. Every output a user sees traces back to a model version, a data snapshot and, where it matters, a human approval.
Finance Tracker
- Problem
- Tracking income and expenses in a spreadsheet works fine until you want to see a trend or compare two periods.
- Approach
- A web application with categorisation, period summaries and cash-flow visualisation — built to be reviewed in a minute rather than an afternoon.
Client work
Paid client work, published with the client's permission.
Chino's Pet Hotel
- Client
- A dog boarding business in Shoreview, Minnesota.
- Delivered
- A bilingual lead-generation site, built so an audience split between English and Spanish reaches the booking form without friction either way.
How I work
Four steps, no surprises.
Diagnostic call
Thirty minutes. I get the problem, the context and what "solved" actually means for you. If I'm not the right person for it, I'll say so on the call.
Scoped proposal
What gets built, what doesn't, how long it takes and what it costs. No ambiguity and no fine print.
Built in increments
Working pieces you can use and judge as we go. I don't disappear for three months and come back with something that wasn't what you asked for.
Handover with documentation
The code, the documentation and the walkthrough. What I build you can maintain without me — that's part of the deliverable, not an add-on.
About
A financial engineer who codes, not a coder who read about finance.
I trained as a financial engineer and I've spent more than ten years working in risk and data science. That's the work I was doing long before I started building software for other people.
The difference matters more than it sounds. Almost any developer can build an application. Very few understand what a reconciliation actually involves, why a projected cash flow breaks, how credit risk behaves across a real portfolio, or what separates a useful forecast from a decorative one.
When the problem is financial or operational, I'm not learning the domain while I build it — I already know it. That means fewer meetings spent explaining your business to me, fewer wrong assumptions, and systems that solve the actual problem rather than the one it looked like from the outside.
10+Years in risk & data
3Markets · Colombia · US · UK
Contact