I build and run the systems, not just the reports. The ones a business uses every day come first here; the analytics fundamentals behind the judgement come last.
Production systems, data and BI work, my own products, and the analytics fundamentals behind the judgement in everything else.
Production systems
Live systems with real clients working on top of them: multichannel inventory, warehouse operations, sales agents, quoting, B2B platforms and client sites.
Three inventories that never talked to each other, unified into a central ledger that decides what stock each sales channel shows — and never oversells.
796 listings audited and an append-only ledger with 1,742 movements.
An offline-first PWA with event sourcing to verify shipments, count stock by location, and replay the full history whenever the numbers disagree.
A replay of 3,536 real events proved the correction was right.
A WhatsApp sales and support engine powered by local models: not a single customer conversation leaves for an external AI service.
Chat, embeddings and voice transcription, all inside my own server.
The customer opens a link, picks occupancy and add-ons, and sees the total update live. What they never see — cost, margin, supplier — is protected by a guard that runs on every response.
A function inspects every public response and throws if it finds a margin.
Approved agencies quote, book and pay for travel experiences. Tenant isolation lives in the database, not in application code.
Row Level Security enabled in PostgreSQL and a test that blocks supplier leaks.
Members, payments, QR check-in, overdue tracking and WhatsApp reminders in an installable app the owner runs without technical help.
v1.5.0 in production, with the running version verifiable from the service itself.
A multi-tenant management system for an inbound travel agency. An automated test cycle found routes reachable by direct URL; today they return 404 by role.
The finding did not come from an external audit: it came from the project's own E2E suite.
Attendance operations system combining WhatsApp capture, a NestJS backend, a React/Capacitor app, and audit-ready traceability for supervisors.
Unifies mobile operations, back office, and notifications inside a single auditable flow.
An online shop and an installable admin panel for a hot sauce producer, with push notifications and a handover designed so the owner does not depend on me.
Delivered with a video tutorial and a ready-to-send WhatsApp message.
A two-language site for a café in Oklahoma City, deliberately minimal: two dependencies, a backup on every save, and a switch instead of a delete button.
«Flip the switch, don't delete the product»: whatever disappears can come back.
Automates invoicing of motorway toll receipts with optical reading and browser automation, and stops when the portal asks for a human.
It does not solve CAPTCHAs or bypass protections: it flags the batch as «needs review».
Data & BI
Data work proper: analytical models over real history, accounting migrations with row-level traceability, local document analysis and measured audits.
A star schema built on a retailer's historical exports: 56,739 transactions reconstructed, and an honest report on what the data cannot answer.
The pipeline reports that only 1.97% of sales value has an identified customer.
An admin panel said «published» while saving to the browser. Around it, 858 MB of PDFs and a homepage that took minutes to load over mobile data.
Headless Chrome at 4 widths; recommended phase 1: one day of work.
Migrating 28 months of spreadsheets into a single accounting ledger, with every record traceable back to its source sheet and row.
An impossible date is not forced into a valid one: it is flagged for human review.
Processes bank statement PDFs entirely on the machine and computes an operational financial health index. No data leaves.
A bank statement is about as private as data gets; here it does not travel.
Interactive Streamlit dashboard to explore Oasis' reunion tour by city, date, estimated attendance, estimated gross, and setlist variants.
41 concerts, 17 cities, and 11 countries condensed into a visual experience backed by public band images.
My own products
What does not come from a client brief: my own products, open demos and cultural pieces I keep running myself, infrastructure and maintenance included.
Thirteen phases covering a convention bureau's full cycle: from prospect to bid, from won event to QR-based evaluation and the executive dashboard.
Public live demo, 67 green tests and 23 design documents.
Digital product focused on speaking-first learning, operational automation, and product experience for English education.
Shows product thinking, operating model, and iteration capability.
Conversational analytics product that turns business questions into safe SQL, charts, insights, and downloadable reports.
Connects natural language, analysis, and visualization in a single self-service workflow.
Mobile-first cultural app connecting mural, map, QR scanning, and audio to activate Copainalá's public art through a published PWA.
Turns a physical cultural route into a navigable digital product ready for public use and promotion.
A tourism information system whose scope comes from a law: every article of the decree maps to a module and to the evidence that satisfies it.
An article → module → evidence matrix, so the scope holds up to whoever asks.
Analytics fundamentals
Training cases built on public or practice data: churn, applied SQL, experimentation and visualisation. They are not client projects and do not pretend to be; they hold up the analytical judgement behind the rest.
Flagship portfolio case that turns a large trending-video dataset into a clear visual narrative across category, region, and time.
339,990 records, 18 categories, and 5 regions condensed into a public dashboard ready for stakeholders.
Analytical case focused on identifying cancellation drivers and surfacing retention and efficiency opportunities in telecom.
Prioritizes churn drivers and high-risk segments.
Advanced SQL case answering acquisition, monetization, and retention questions with reproducible, auditable queries.
ARPU, LTV, and churn computed with CTEs, joins, and window functions.
Hypothesis validation across digital flows using Welch's t-test and assumption checks to support conversion decisions.
Compares A/B variants and separates statistical signal from noise.
Lightweight app for exploring vehicle price, mileage, and condition with live filters and interactive charts.
Turns exploratory analysis into a browsable tool for non-technical users.