Senior Test Automation Engineer · AI QA & Test Strategy · M.Sc. in Artificial Intelligence 📍 Medellín, Colombia · 🌎 Open to remote work · 🗣️ Spanish (native), English

I've spent 13+ years making software trustworthy: 3 as a developer and 10 in quality engineering. Today I'm bringing that mindset to AI. I believe a model is only as good as the way we evaluate it, so I focus on testing, measuring and explaining ML systems, not just building them.

• Test automation & quality strategy for data-heavy and integration-heavy systems (Python/pytest, Java/TestNG, Cypress, Robot Framework, POM).
• Data & ETL quality: source-to-target validation, reconciliation and quality gates on Azure and AWS pipelines.
• ML model evaluation: ROC-AUC, PR-AUC, calibration, robustness under distribution shift and explainability with SHAP.
• AI-assisted QA tooling: I build agent skills (Claude Code, Kiro) that my team uses in daily delivery, like xray-sync, which detects missing test scenarios during suite migrations and creates them in Xray automatically.

At Globant I've worked across film marketing, financial systems and video-on-demand streaming. I've been Technical Lead, release Gatekeeper and mentor for other QA engineers.

• 🎓 Interpretable Deep Learning for SME Credit Scoring (M.Sc. thesis): a reproducible benchmarking pipeline comparing TabNet, XGBoost and logistic regression, with multi-criteria evaluation and three layers of interpretability. Core idea: interpretability is a property of the evaluation pipeline, not only of the model.
• 🤖 TRAZO, Factored AI & Data Hackathon 2026: a governed AI agent for disputed card charges whose autonomy is set per intent from Wilson score bounds on the human override rate. It only acts alone when the evidence says it's reliable.
• 🧪 Evaluation-Sensitive Behavioral Modulation in Frontier LLMs: a preprint on measuring whether LLMs behave differently when they know they're being evaluated.
• 📝 ML blog: beginner-friendly Machine Learning with Python, and notes for anyone moving from QA into AI.

• Fair access to capital: transparent, explainable credit models that help small businesses get financing banks can trust.
• Explainable & Responsible AI: models people can audit and understand.
• AI evaluation & red teaming: stress-testing models and agents to see if they behave the way they should.
• Causality and better ways for AI systems to reason about cause and effect.

I'm growing from quality engineering into AI / ML engineering and AI evaluation. My goal is to work on hard problems with real social impact, where rigorous testing and explainability are part of the design from day one, not something added at the end.

Happy to talk about AI evaluation, explainable ML, test strategy or anyone's journey from QA into AI.