Operations · Internal tooling · Data & AI automation
I'm an operations person who actually likes people — and I'd rather build the fix than file another ticket.
As COO of a growing credit-restoration & financial-education company, I own the operating model — and build the tools, data pipelines, and compliance behind it. Employer, brand and client data are anonymized; every screenshot uses synthetic data.
Links every finding to immutable evidence, recommends only fact-supported methods from a 21-method engine, generates bureau letter packages, tracks legal deadlines, and keeps a hash-chained audit trail — with client PII kept out of storage by design.
60+ Python scripts dedupe, migrate, reconcile and QC a live CRM; every destructive operation writes a change log and a backup. Designed a write-once attribution field and backfilled 865 client records so commissions roll up permanently.
A family of browser-based tools that parse a report in-memory and produce an instant plan. When printed totals disagreed with the detail, the parser counts the red-X marks per bureau instead of trusting the total.
A pipeline that consolidates CRM revenue, bank/card, payroll and funding into branded dashboards — 2,089 payments across 33 months, every figure traceable to a named source file.
Designed a three-tier curriculum (10 courses / 86 lessons) and a searchable written knowledge base with a build-time PII guard that strips any link exposing client identity.
Authored the internal knowledge base & SOPs, communication standards and escalation matrices, plus custom AI skills (a brand director and a compliant-comms writer) so a remote bilingual team produces on-brand, compliant work.
10+ years across workforce management, real-time operations, training, and operations leadership.
Self-taught data-systems & automation work, built against real projects — sectors and clients withheld for confidentiality. Each is problem → method → outcome, with an honest “what I’d do differently.”
Seven authorities that disagreed, sixty-two entries that didn't belong, and the audit that found what my verification had missed.
BOne contract, two implementationsRead →Written twice on purpose, because no single fetcher could reach every source — and the coordination problem I documented instead of fixing.
CThe independent recountRead →Building something whose only job was to disagree with me — and the governing body whose own current list answered a different question.
DBuilt to need no tuningRead →Composed artwork from arbitrary photographs with no human adjustment — five rejected passes, every constant traced to the failure it replaced.
EPer-field authorityRead →Two systems that disagreed, where neither could simply win — and the duplicate hiding behind a guardian's contact details.
FThe expensive directionRead →Automated image search across hundreds of named minors, where the failure isn't finding nothing — it's confidently finding the wrong person.
GThe limitations sectionRead →Five near-identical multi-hour files named only by timestamp — identified from evidence inside them, delivered with an honest account of what they are not.
HSeven monthsRead →What actually changed, the mistake I made three separate times, and an honest list of what this period doesn't prove.
◆The working recordReceipts →The evidence behind the claims: real code, real schemas, and a file inventory pulled straight off my drive.
I make workflows less painful — sometimes with a better process, sometimes with a spreadsheet, and sometimes with a small app I build and ship myself.
I build pipelines that turn messy, contradictory sources into one record you can trust — and keep it honest with verification, matching, and provenance.
I turn "how it's done" out of someone's head and into clear guides, onboarding, and training that people can actually follow.
I'm genuinely a people person — warm with customers and teammates, and steady when things get messy. That's usually the part that holds everything together.
A few rules I actually live by — each one learned by getting something wrong first.
Sources named, rules fixed, output decided — before anything runs. A vague brief just gives you plausible output nobody can check.
Nothing moves on a single pass. A second, independent check has to agree — and if it doesn't, the work stops.
I'd rather flag "we don't know this yet" than fill it with a value that looks right and quietly isn't.
The limitations note is the one nobody asks for — and the reason everything else can be trusted.
American School of Management.
Universidad de San Carlos de Guatemala · 2010–2013.
Certified proficient — comfortable leading client calls and written work in English.
Applying AI tooling to real operations and productivity.
Certified in training-needs analysis and enablement.
CRM, Inbound & related HubSpot platform certifications.
Power BI · Tableau · Excel · WFM (NICE, Verint, Genesys, IEX, Calabrio) · Salesforce · Zendesk · Zapier · Make · Power Automate · Python · AI tools.
Employment references and documentation available on request.
Hi — I'm Carolina. I like the parts of a business most people would rather avoid: the messy handoffs, the spreadsheet nobody wants to own, the process that only lives in one person's head.
I've spent my career in operations and customer-facing roles, and the thread that runs through all of it is people — making their day a little easier, their tools a little simpler, their first weeks a lot less overwhelming. When something's repetitive, I'll build a small tool or write the guide so nobody has to suffer through it twice.
In the last year I also taught myself to build proper data systems — pipelines that reconcile thousands of records, check their own work, and flag what they can't be sure of. It's some of the work I'm proudest of, and it grew straight out of the same instinct: make the messy thing trustworthy.
I work remotely from Guatemala, in English and Spanish, and I'm just as comfortable running a call as I am writing an SOP or shipping a little app.