Luis – Workflow Automation, SQL, Webhooks

Sole owner of technology, automation, and business intelligence for a bank's commercial directorate. Designed and delivered AI-powered internal tools that automate manual processes, partnering directly with business stakeholders and owning solutions end to end.

• Developed 10+ AI-powered n8n workflows, including a company-dossier tool that cross-references 6+ data sources from a tax ID and emails an AI-synthesized report in ~2 minutes (down from ~40 minutes of manual research);
• Implemented and solely administer the company's first CRM (70+ active users) with automated sales funnels and SQL Server API integration; co-designed the first formal outbound SDR prospecting flow;
• Created a standalone email-marketing platform (~15K recipients) with SQL-based segmentation, open/click tracking, and an analytics dashboard, replacing paid third-party tools;
• Streamlined a ~200/month client-portfolio migration workflow, cutting lead time from ~2 days to under 4 hours with conditional approvals and a full audit trail;
• Orchestrated a cross-department fee-waiver approval workflow (~50/month) with threshold-based routing and compliance traceability.

A unified platform that brings together four motorcycle-business verticals (e-commerce, barbershop, riding school, and tuning workshop) into a single web and mobile application, combining online sales, service bookings, event registration, customer engagement, and back-office operations.

• Built a full e-commerce flow with a catalog synced to an ERP (Bling) over OAuth + webhooks — a paid order is created automatically in the ERP — with Pix/card/boleto payments via MercadoPago and shipping calculated by postal code;
• Built a 4-step booking system with conflict validation and a multi-step event-registration flow with progressive pricing tiers and automatic seat control;
• Delivered a 24-page admin panel with role-based access (owner/admin/staff) and financial dashboards;
• Shipped web and native iOS/Android from one React codebase via Capacitor, automating iOS builds with Codemagic (Mac Mini M2 -> TestFlight) and configuring deep links;
• Wrote 76 versioned SQL migrations with row-level security and generated TypeScript types; covered checkout with Vitest + Playwright across 14 Edge Functions.

An AI-powered nutrition tracking platform that replaces manual food logging with natural voice or text conversations, automatically converting meals into structured nutritional data. Available as a PWA and native iOS/Android app, with progress tracking, diet plans, nutritionist–patient collaboration, and subscription management.

• Designed and shipped the end-to-end product solo, from a 38+ table PostgreSQL schema with row-level security to native store releases;
• Built the AI logging pipeline (198 n8n nodes): Whisper transcription -> Claude Haiku food extraction -> macro estimation, persisting meals through Supabase Edge Functions;
• Cut AI cost and latency with a two-layer food cache (a normalized fuzzy-match lookup backed by on-demand Claude Haiku estimation that persists results), choosing Haiku deliberately to keep per-request cost low;
• Implemented an offline-first message queue persisted in localStorage that syncs automatically on reconnect;
• Secured the backend in layers: PKCE auth, RLS on every table, JWT-validated webhooks, per-origin CORS whitelist, rate limiting across 14 Edge Functions, PII stripped from logs;
• Set up Codemagic CI/CD for both platforms (signed Android AAB, iOS entitlements, automated TestFlight upload); added a built-in RAG support chatbot.

A multi-tenant digital concierge platform for residential buildings that replaces manual delivery notifications with real-time resident communication via WhatsApp, push notifications, and an in-app digital intercom. The platform also includes AI-powered document search, building-wide announcements, and resident management tools.

• Built the automation layer with n8n webhooks driving WhatsApp delivery (Evolution API) and Firebase Cloud Messaging push, logging every send for auditability;
• Engineered a real-time "digital intercom" using Supabase Realtime (postgres_changes) so the concierge sees the resident's response without polling;
• Designed multi-tenant isolation with a building ID on every table and RLS helper functions, supporting many buildings from one backend;
• Added a RAG chat over building documents (OpenAI embeddings + Claude via n8n) rendered as Markdown;
• Led the architecture migration from localStorage + Google Sheets to Supabase Auth + PostgreSQL and documented a security review;
• Shipped web (Vercel) and native iOS/Android (Capacitor + Codemagic), dynamically importing native plugins to avoid crashes on web builds.

A digital operations platform for a cat café and its partner animal-rescue NGO, combining event management, cat adoption, reservations, payments, and business administration in a single system. Replaced manual WhatsApp and spreadsheet-based workflows with integrated customer and operational tools.

• Implemented a 4-step event booking flow (details → form → PIX payment with receipt upload → WhatsApp confirmation) with temporary seat holds during checkout and an event waitlist;
• Developed an adoption showcase with individual cat profile pages and an animated homepage counter modeled on MAX(sort_order) instead of COUNT, ensuring accuracy when cats are returned;
• Designed a dual data source (Supabase ↔ legacy Google Sheets via Apps Script) switchable by environment variable, allowing the non-technical owner to continue managing events from a familiar spreadsheet;
• Created a protected admin panel (Supabase Auth) with a customizable dashboard (occupancy heatmap, live occupancy, top clients, activity feed, system status), plus recurring events, manual bookings, batch notifications, schedule overrides, client history, and reports;
• Introduced a job-application workflow with admin management and proxied payment-receipt uploads through a Supabase Edge Function to keep the image-host API key server-side;
• Conducted eight rounds of end-to-end audits covering security, accessibility, and critical bugs, and migrated production hosting from Vercel to Cloudflare Pages (+ Pages Functions).

A multi-tenant corporate knowledge platform that transforms uploaded documents into a searchable AI assistant using OCR, embeddings, and RAG. Users receive cited answers linked directly to source documents, with permissions enforced at both application and retrieval layers.

• Architected the end-to-end RAG pipeline, orchestrating ingestion (OCR -> chunking -> embeddings) and querying through external n8n workflows to decouple processing from the frontend;
• Implemented tag-based access control enforced twice — through Supabase RLS and inside a custom pgvector RPC with an admin bypass;
• Built a quantitative RAG evaluation harness in Python with RAGAS (faithfulness, answer relevancy, context precision/recall) to track quality regressions instead of relying on subjective testing;
• Ran a formal security audit, identifying 6 vulnerabilities and implementing fixes for prompt injection and an admin-bypass flaw;
• Delivered real-time document processing status via Supabase Realtime for uploads up to 150MB; built RBAC and a 27-event audit log with CSV export.