Visual evidence
Architecture, workflows & implementation evidence
Architecture diagrams, workflow screenshots, catalog examples, before/after results, and implementation evidence from the BAYNAO platform.
Engineering evolution
From manual commerce to AI-powered operations
BAYNAO did not begin as an AI project. The current architecture is the third generation of a business platform that has been progressively rebuilt as operational needs, infrastructure, and available technology changed.
Custom PHP Commerce
Semi-dedicated Linux · PHP · MySQL
Custom storefront, catalog and administrative tools. Product updates, quotations, orders and invoicing were managed through the PHP system, while WhatsApp customer conversations were handled manually.
Modernization & Rule-Based Automation
AWS · Modern PHP · Composer · Libraries
The platform moved to AWS and the PHP application was modernized. A deterministic WhatsApp bot automated structured quotation and order intake, while open-ended customer questions still required human assistance.
AI-Powered Automation
Local Linux · Cloudflare · APIs · n8n · AI · R2
The current platform combines the original business data with custom APIs, workflow orchestration, AI, multimodal processing, automated catalog maintenance, QA systems and new interaction channels.
Generation 01
Custom PHP commerce & manual WhatsApp service
The first BAYNAO generation was a custom PHP/MySQL e-commerce platform hosted on a semi-dedicated Linux server. I built and maintained the public commerce experience together with the internal administrative functionality used to manage the business.
- Custom PHP e-commerce website and product catalog.
- MySQL-backed product and business data.
- PHP administration for catalog updates and operational records.
- Management of quotations, orders and invoices.
- Customer service and product questions handled personally through WhatsApp.
The important limitation was not the commerce platform itself: the sales conversation still depended heavily on manual attention.
Generation 02
AWS modernization & rule-based WhatsApp automation
The second generation combined an infrastructure/application modernization with the first major automation of the customer sales process. BAYNAO was migrated to AWS, PHP was updated to a more modern environment, and Composer plus additional libraries were introduced.
I also built a rule-based WhatsApp bot without AI. Instead of understanding unrestricted natural language, it guided the customer through a deterministic sequence of questions required to prepare a quotation or order.
- Collected structured customer and transaction information.
- Applied location-based business logic to delivery/pricing decisions.
- Prepared quotations and orders.
- Generated the resulting PDF document.
- Returned the PDF to the customer through WhatsApp.
Generation 03
Local infrastructure, Cloudflare, APIs, n8n & AI
The current generation keeps the accumulated business logic and catalog data, but changes how systems interact with it. BAYNAO now runs through local Linux infrastructure exposed and protected through Cloudflare, with custom APIs allowing automation and AI workflows to consume operational data programmatically.
This generation moves beyond a fixed questionnaire. Customers can communicate naturally, while the system determines the type of request, retrieves business data, maintains relevant context and routes work to specialized workflows.
Current system
AI-assisted sales and customer operations
- Intent routing: product search, technical questions, quotations, orders, FAQs, customer data and human assistance.
- Natural-language product search: customer requests are translated into structured searches against BAYNAO's product data.
- 2,000+ product catalog: prices, availability, specifications, alternatives, relationships and imagery can be reused by automation.
- Conversation context: follow-up messages can continue previously discussed products, quantities, quotations and pending orders.
- Multimodal processing: text, images, audio, PDFs, documents and URLs can be normalized before routing.
- Quotation & order workflows: customer, product, quantity, delivery, payment and transaction data are structured and validated before proceeding.
- Human escalation: automation can hand work back to a person when the request requires review or the system lacks reliable information.
Supporting engineering systems
Automation around the core platform
The current BAYNAO ecosystem also includes specialized systems that maintain, enrich and validate the data consumed by customer-facing automation.
- Custom PHP Product API: reusable access to product and business data for workflows and AI systems.
- AI Product Builder / Catalog Enrichment: researches products by SKU on the web, combines evidence with database data, generates structured catalog/SEO content and applies deterministic protections before updates.
- AI Image Updater & Hash Engine: detects products without images, searches by SKU, downloads and validates assets, organizes catalog imagery, generates perceptual hashes and publishes to Cloudflare R2.
- AI Image Quality Auditor: provides a second QA layer to inspect and reinforce the image-updater results.
- API Quality Auditor: production regression and consistency testing for the commerce API.
- Warranty Browser Automation: automates repetitive warranty lookups in an external portal and synchronizes results with internal records.
- Voice AI System: connects n8n with a local Issabel telephony server for AI-assisted voice workflows.
Engineering significance
The same business platform, progressively redesigned
The value of BAYNAO is not only the current AI layer. It demonstrates a long-running engineering process: understanding an operating business, building the original software, modernizing infrastructure, automating deterministic processes, exposing reusable data through APIs, and finally introducing AI where natural language and flexible reasoning add value.