LUIS RODRIGUEZ RIVERA / AI-AUGMENTED FULL-STACK ENGINEER
PRODUCT DELIVERY + EVALUATION-DRIVEN DEVELOPMENT
I turn foundation models into a controlled engineering workflow: define the goal, generate a candidate, challenge it with role-specific reviewers, correct the findings, verify the system, and make the final human call.
STAGE 01 / PROOF OF WORK
Three working systems that show how I think: public safety, operational awareness, and evaluation-driven AI delivery. No fictional metrics—open the products and test them.
LIVE / PUBLIC SAFETY
MISSION 01 · FEATURED
Official ocean data, translated into a decision-ready experience.
A bilingual public-safety platform for 98 Puerto Rico beaches. It combines NWS alerts, NOAA tides, buoy observations, hazard profiles, geolocation, and an interactive map—while clearly exposing when a live source cannot be verified.
Next.js · React · TypeScript · Leaflet · NWS / NOAA · PWA
LIVE / OPERATIONS
MISSION 02
The dashboard I built because status hunting was slowing me down.
A mission-control view that consolidates operational status and incident context across 14 AI, hosting, database, and developer services. Designed to be glanced at, not studied.
Next.js · TypeScript · Status APIs · Monitoring UX
MISSION 03
See how AI-generated work earns acceptance.
A working acceptance loop for agent-written software. GPT‑5.6 returns schema-bound domain, security, product, and test findings; ProofLoop turns them into traceable correction and regression evidence while keeping final authority human.
OpenAI Responses API · JavaScript · Zod · Netlify Functions
BONUS STAGE / CREATIVE RANGE
COD Zombie SlugFan experiment · credited sprites and physics
PLAY ↗
Eval FlightInteractive AI evaluation mission · judges, evidence and human acceptance
LAUNCH →
STAGE 02 / THE DIFFERENTIATOR
I use coding agents every day, but never treat their first answer as the finished product. The advantage comes from the control loop around the model.
AI-generated code is an untrusted contribution until deterministic checks and human judgment confirm it.
I bound work with explicit goals, context, constraints, and completion criteria so the model optimizes for the real task—not merely plausible code.
Role-specific reviewers inspect architecture, security, domain correctness, tenant isolation, frontend behavior, and test quality from different angles.
I choose Claude, Codex/ChatGPT, or Gemini based on the task, protect the context window, and spend tokens where another pass reduces real risk.
I inspect the output, run the system, challenge assumptions, and remain accountable. Judges help find defects; they do not replace evidence.
STAGE 03 / CAMPAIGN HISTORY
Full-stack delivery across AI evaluation, payments, healthcare, document workflows, public-sector systems, and quality engineering.
Handshake AI · Contract
Business Computer POS
Vaco by Highspring · Contract
Puerto Rico Department of Health · Contract
Truenorth Corporation · Contract
Require Technology · Contract
FINAL STAGE / LET'S BUILD
I’m interested in applied AI, AI product engineering, and full-stack roles where shipping reliable software matters more than performing buzzwords.