Lives in the workflow
Direct links open the relevant evidence without asking someone to learn a new portfolio structure first.
AI PRODUCT ENGINEER · AVAILABLE FOR BOUNDED TRIALS
Working software for
messy operations.
I turn ambiguous AI ideas, manual workflows, and incomplete products into systems people can inspect, test, and use.
$60K+earned on Upwork
5,600+client hours delivered
END–TO–ENDproduct, AI, data, QA, deployment
SELECTED PRODUCT EVIDENCE
Don’t take the positioning on faith.
Every item below is either an interactive synthetic demonstration or a case study grounded in inspectable work.
01
INTERACTIVE DEMO
AI Operations for Sales & Support
A synthetic WhatsApp-first operations console where AI drafts answers, cites business knowledge, qualifies intent, and routes uncertain cases to a human.
AI workflow · human handoff · evaluation
02
INTERACTIVE DEMO
Operational Memory
A source-citable coordination system that turns scattered project activity into decisions, blockers, actions, and reviewable recommendations.
Citations · approval gates · audit trail
03
PUBLIC CASE STUDY
Signal Conversation Export
A local-first macOS tool that exports Signal Desktop conversations to structured JSON, readable Markdown, and optional media bundles.
Python · packaging · 6 verified tests
04
PUBLIC CASE STUDY
Crisis Damage Intelligence
A bilingual, static-first geospatial evidence package that keeps official source data separate from model-assisted review.
Geospatial · data QA · deployment
PROOF THAT TRAVELS
The output should be useful after the pitch ends.
Each demo and case study is a standalone object: open it from a proposal, test it in the browser, forward it to a teammate, or use it to frame a real workflow.
Designed to be forwarded
Every detail page can copy or share its own URL and still makes sense when it arrives without a sales call.
Owned beyond a platform
The demos connect back to my writing and identity at luisrosal.dev, while Upwork remains the contract workflow.
THE OPERATING MODEL
Build the smallest thing that can answer a real question.
Frame
Name the decision, user, risk, and acceptance threshold before expanding scope.
Ship
Deliver one complete vertical workflow across interface, data, AI, and safeguards.
Observe
Instrument the critical action and collect failures, friction, and user behavior.
Decide
Continue, iterate, or stop from evidence—then leave a clean technical handoff.
LOW-RISK FIRST ENGAGEMENT
A five-day product evidence trial.
Give me one rough idea, workflow, repository, or recurring failure. I’ll define the smallest useful test, ship a working vertical slice, instrument the critical behavior, and deliver a clear continue/iterate/stop recommendation.
Working buildAcceptance checksEvidence briefNext-step plan
Continue the conversation on Upwork ↗