One person, an agent fleet,
and the judgment in the middle.

The arc

My path has been a steady climb toward owning bigger, more strategic delivery:

• Healthcare compliance — led compliance initiatives at Humana (Medicare Advantage), automating audit workflows and data validation under HIPAA and Medicare Parts C and D.
• Quality engineering — built automated test frameworks and data-model validation: the discipline of proving a system does what it claims before it ships.
• Data engineering — designed and built end-to-end pipelines, BI, and API integrations for a major public utility (PRASA via TrueNorth), and led data-governance work across its analytics.
• AI and automation leadership at Optum (UnitedHealth Group) — I now lead projects for automation and AI integration, including LLM solutions, and the migration of legacy systems into modern, ML/AI-driven platforms. Two in-house product implementations are in production.

The throughline: compliance taught me what “correct” has to mean in a regulated, high-stakes domain; quality engineering taught me how to verify it; data engineering and AI leadership are where I now build and direct the systems.

Where I work, and who I build for

I’m based in Puerto Rico and build for the PR market first — bilingual (EN/ES) products, local data sources, and the Act 60 context that matters to business clients here. The center of gravity in my work is leading and growing the work and the people, and getting hands-on when it matters.

The meta-story: one senior person, an agent fleet, and the judgment in the middle

Here’s the part that’s harder to see from a resume. Everything in my personal portfolio — a live real-estate marketplace, a procurement-intelligence product spanning six government data systems, and a business OS running on AWS — was designed, built, deployed, and operated by one person directing AI agents as the implementation layer.

The agents are the hands. The human owns the parts that don’t delegate: architecture, judgment, and verification. That last one is where most of the work lives. A few examples from these systems:

• A 50-agent QA fleet stress-testing the FT-OS business OS.
• A 44-agent gap analysis on the LicitaPR data backend.
• An authorized Nidopr security audit — 21 findings, posture strong.
• A 15-skeptic adversarial test that killed a tempting-but-false finding before it could ship — exactly the kind of plausible result a less skeptical process would have published.

That last one is the whole point. Agents will happily generate a confident, wrong answer. The job is to be the person who refuses to ship it until it has been tested hard enough to break it.

The honest line

AI doesn’t replace the engineer. It lets one senior person lead and deliver what used to take a team — while still owning the architecture, the judgment calls, and the verification that decide whether the work is actually true.

That’s the skill I’m bringing.