A CV hides the structure. A graph exposes it — but most graphs are software projects about the data, not arguments made with the data. This one is hand-composed. Every mark is where it is for a reason. Time flows diagonally from Bogotá to London. The present is the largest word on the page.
Five years of rigorous engineering training at Universidad de los Andes. Foundations in algorithms, software architecture, distributed systems, and applied mathematics — the groundwork that everything since is built on.
Undergraduate thesis applying machine learning to Formula 1 race-outcome prediction. A rigorous exercise in custom dataset construction, feature engineering, and evaluation across a 70-year historical corpus.
Early industry experience inside one of Latin America's largest banks. Delivery discipline, cloud fundamentals, and systems thinking inside a high-stakes financial environment.
Fast-moving startup. Designed the core serverless infrastructure from scratch; shipped LLM-powered travel experiences before they were mainstream.
Postgraduate specialisation in software engineering, applied machine learning, and evaluation methodology. A formal reframing of production engineering through the lens of research-grade rigour.
Thesis work on performance-regression detection in high-performance computing — a research problem at the intersection of deep learning, systems engineering, and reproducible science.
Driving the AI feature roadmap at a private-equity-focused software platform. Architecting and shipping production ML systems that reduce analyst workload and accelerate investment decisions.
End-to-end extraction system that turns unstructured financial documents into structured data — KPIs, targets, metrics. Deployed on GCP and integrated directly into Untap's workflows.
AI assistant integrated into Untap's platform — helps analysts query portfolio data in natural language, generates memos, surfaces intelligent insights. Built on a custom MCP server connecting LLMs to proprietary data sources.
SwiftUI iOS habit tracker with an AI coaching layer powered by the Claude API. Explores the design space between structured self-tracking and context-aware personal intelligence.
This website — a living experiment in interaction design, performance trade-offs, and visual storytelling. Built to the same standards as production work.
The composition you're reading. Hand-placed typographic nodes, drawn connections, no force simulation. A career told as a map, not a timeline.
A production-grade multi-task transformer for football intelligence. Match outcomes, xG modelling, scoreline simulation, player performance, Monte Carlo scenario planning. In progress.
LLM evaluation framework — systematic, reproducible, actionable. Born from first-hand experience with the limitations of ad-hoc evaluation in production AI.