Bardie AI
? Oops! Sorry
The mobile version is not created.
I'm working.
Please visit on desktop version.
Project
Bardie
Date
Tasks
March 2026
Product Design
IT & Sales Strategy
Prompt Engineering
RAG Architecture
PHP Backend
AI Fine-Tuning
Vibe Coding
Link
Doc
Map
An Engineering Kerning LABS® Case Study
Engineering a production-ready AI sales agent, from zero to deployment in 7 days. I built Bardie, a proprietary RAG system and multi-tenant PHP infrastructure to prove that generative AI can autonomously execute complex sales methodologies.
#1 The Vision
Most AI systems are built to answer questions. Bardie was built to close deals. The vision behind this project was to close the gap between technical automation and human sales psychology — transforming a generative AI from a passive support tool into a high-performance commercial agent capable of executing complex methodologies like ZMOT and Three Gaps across multiple Buyer Personas simultaneously.
The result is a fully deployed Generative AI system (RAG) purpose-built for pre-booking automation, taken from concept to functional product in a 7-day development cycle. Built on a validated sales core powered by Groq, an asynchronous messaging architecture via SMTP and Cron Jobs, and a scalable multi-instance infrastructure, Bardie manages complex sales flows and real-time data synthesis natively — integrating with email protocols or chatbots, and architected from day one for final scaling on Gemini 2.5.
#2 Strategic Core
The core challenge was engineering a sales agent capable of executing ZMOT and Three Gaps methodologies across 7 distinct Buyer Personas with zero hallucinations and full contextual precision. This required building a RAG knowledge base that could handle complex, multi-persona conversations while maintaining strict response accuracy at every touchpoint.
The solution was architected around a proprietary PHP core (Instille Multi), ultra-fast inference via Groq, and a fully asynchronous messaging layer powered by SMTP and Cron Jobs. The result was a complete operational ecosystem including AI log dashboards, system mapping, and technical documentation shipped end-to-end in 7 days.
- Inference Engine & Scalability: Initial implementation with Groq (LPU) for minimum latency in the PoC phase, with a decoupled architecture ready for migration to Gemini 2.5.
- Proprietary Middleware (Instille Multi): Orchestrator built in PHP with a multi-instance architecture, enabling the deployment of multiple isolated environments (motorcycles, boats, etc.) under a single logical core.
- RAG Architecture (Retrieval-Augmented Generation): Dynamic knowledge base with hierarchical dependency injection to ensure precise responses grounded in 7 Buyer Persona archetypes.
- Async Communications Management: Integration of SMTP protocols and automation via Cron Jobs for managing sales threads and automated responses 24/7.
- Monitoring & Control (System Dashboard): Centralized control panel for supervising AI logs, SMTP server status, and real-time traceability of conversation flows.
- System Mapping & Technical Documentation: Fully documented architecture, including logical process mapping, decision flows, and implementation guides for product scaling.
- Frontend & Conversion Assets: Deployment of high-speed Landing Pages and end-user Dashboards focused on pre-sale metrics visualization and lead capture.
#3 Work Process
The Challenge: Can a production-ready AI sales agent capable of executing complex selling methodologies, be built and deployed in just 7 days? This project was a technical sprint to transform a generative AI concept into a fully functional pre-booking engine. Starting from zero, the entire stack was architected and shipped: from the RAG knowledge base and multi-tenant PHP middleware to async messaging infrastructure and real-time monitoring dashboards.
Every decision was made to eliminate the gap between sales psychology and technical automation deploying a system that thinks, responds, and converts 24/7 without human intervention.
- 01. Prompt Engineering & Sales Logic: Defined the business logic framework based on ZMOT and Three Gaps methodologies, modeling the RAG knowledge base for 7 Buyer Personas — zero hallucinations by design.
- 02. «Instille Multi» Architecture: Evolved the injector into a multi-tenant PHP infrastructure, enabling instant deployment of isolated instances with pre-configured cores under a single logical engine.
- 03. Async Implementation (PoC): Integrated Groq for ultra-fast LPU inference and built the messaging engine via SMTP and Cron Jobs, enabling autonomous management of sales flows 24/7.
- 04. Automatic RAG Optimization: Deployed a synthesis module that processes large-scale product datasets and automatically converts them into AI-optimized knowledge structures with hierarchical dependency injection.
- 05. Fine-Tuning & Gemini Scaling: Refined Bardie’s core through iterative optimization of sales processes and email threads. This phase consolidates the methodology to transfer the trained «brain» from the PoC into the final production infrastructure on Gemini 2.5, ensuring superior performance in the commercial release.
#4 Results
Bardie compressed what typically takes months into a 7-day delivery cycle — from blank canvas to a fully operational lead-capture system. The pre-booking engine runs autonomously, managing 3 to 4-email sales sequences with full contextual coherence and zero human intervention.
Built on a multi-tenant architecture, the platform is designed to scale across unlimited business instances, each with independent product catalogs and fully customizable sales logic. The PoC was validated with real client interactions email sequences delivered, responded to, and converted autonomously, confirming the system’s readiness for full commercial rollout. 🚀😊
ClickConvert
? Oops! Sorry
The mobile version is not created.
I'm working.
Please visit on desktop version.
Project
ClickConvert
Date
Tasks
March 2026
Product Design
IT Strategy
Business Arch.
Branding
Web
ERP
Vibe coding
An Engineering Kerning LABS® Case Study
Designing a high-performance conversion framework from the ground up. I built this proprietary PHP framework and ClickFlow® engine to prove how custom MVC and surgical UX transform hypotheses into high-ROI infrastructures.
#1 The Vision
Most businesses are inadvertently bleeding ROI by relying on «bloatware-heavy» CMS platforms. These generic systems are built for the masses, sacrificing loading speed and user experience for a suite of features that offer zero value to a high-ticket sales funnel. In a world where every millisecond costs money, «standard» is no longer enough.
ClickConvert was born from a radical philosophy: Surgical Precision. It isn’t just a website; it’s a proprietary Full-stack PHP infrastructure engineered to bridge the gap between digital traffic and actual revenue. By applying a strict MVC architecture and «Zero-Bloat» coding, I have stripped away the technical debt of traditional systems to prioritize what truly matters: speed, traceability, and conversion.
With a native 98%+ Lighthouse performance and sub-100ms response times, ClickConvert ensures that the path from a Google Ads click to a qualified lead is frictionless. The vision is to transform technical excellence into a business weapon—moving beyond vanity metrics to focus on a single, uncompromising goal:
The Maximum Conversion.
#2 Strategic Core
ClickConvert isn’t just a website; it’s a dynamic sales funnel. My workflow integrates market-fit research with a proprietary ClickFlow system:
- Step-up Profiling: Instead of long, friction-heavy forms, the framework uses a progressive qualification logic. It captures basic data first and injects complex fields (budget, specs, urgency) as the user engages, ensuring a surgical filter of lead quality.
- Ad-Hoc Zero-Bloat Architecture: Built entirely from scratch in strict PHP MVC, this framework eliminates the heavy dependencies and bloatware of traditional CMS platforms. By guaranteeing load times under 100ms and a consistent Lighthouse score of >98, it minimizes bounce rates, directly lowering the Cost Per Acquisition (CPA) and maximizing Google Ads ROAS.
- Behavioral Fingerprinting & Predictive Scoring: A built-in Business Intelligence engine calculates the real «temperature» of a lead dynamically. It tracks engagement depth—such as the frequency of visits, time spent on case studies, and interaction within a 48-hour window—to assign a dynamic Sales Velocity score. This allows sales teams to prioritize high-intent micro-targets instantly.
- End-to-End Attribution & True ROI Tracking: Moving beyond vanity metrics and simple lead volume, the framework seamlessly syncs ad spend with closed deals. It traces the entire user lifecycle from the first Google Ads click to the final signed contract months later, revealing the net Return on Ad Spend (True ROAS) and the real Lifetime Value (LTV) of each campaign.
- Workzone ERP: A centralized backend that provides total lead traceability, from the first Google Ads click to the final conversion, including real-time interest scoring.
- Automated Nurturing: Integrated SMTP and API logic that «warms up» leads with automated, transactional emails based on user behavior and project status.
#3 Work Process
The Challenge: Can a high-performance CRO framework be conceptualized, branded, and deployed in just two weeks?
This project was a personal challenge to evolve a conversion framework from 2024 into a fully functional MVP, leading the entire cycle from strategic architecture to the final product deployment.
- 01. Time-Boxing & Research: I established a strict 14-day window to move from ideation to a multi-language landing page. The process began by auditing my 2024 research docs and using html.to.design to import validated structures into Figma, allowing for rapid UI prototyping without reinventing the wheel.
- 02. Interaction Architecture: Once the UI flows were defined in Figma, I drafted 5 comprehensive specification documents. These covered every technical touchpoint: dynamic forms, modal logic, interlinking, and UX triggers, ensuring the system’s «brain» was fully mapped before writing a single line of code.
- 03. Vibe Coding & Prompt Engineering: Using Claude as a high-speed development partner, I imported the visual designs and the 5 spec docs. I leveraged «Vibe Coding»—directing the AI to generate a custom «All-in-one» HTML/Tailwind CSS structure, focusing on architectural integrity rather than just syntax.
- 04. Iterative Fine-Tuning: I performed 20 to 30 iterations per page to ensure 1:1 fidelity with the Figma design. This stage was critical to refining JS interactions and polishing the «Kerning» (the details) of the code to reach professional standards.
- 05. Systematization & Backend Logic: I manually configured the internal processing logic (config-send.php, project-form.php, and transactional email templates). This ensured that the AI-generated frontend integrated perfectly with a secure, functional backend.
- 06. Performance Validation: The result is a lightweight, ultra-fast infrastructure. Despite the speed of development, the core architecture achieved a 97+ PageSpeed score, proving that AI-guided code, when strictly directed, can outperform traditional heavy CMS builds.
#4 Results
I am incredibly proud of this milestone: delivering a full-scale digital product in just 14 days. Despite not being a developer by trade, I successfully architected the core ClickFlow® framework, designed the high-conversion landing infrastructure, and engineered a fully functional ERP backend with automated lead nurturing.
This project proves that with ‘Vibe Coding’ and strategic direction, the gap between a complex hypothesis and a production-ready, scalable MVP can be closed in record time. From pure logic to final deployment: it’s all live ?
Explore the final deployment
A high-performance conversion engine, delivered in record time.
Profe por horas
? Oops! Sorry
The mobile version is not created.
I'm working.
Please visit on desktop version.
A new platform in which you choose the learning, time and budget.
In 2018 I studied a Master of UX/UI at EsDesign Barcelona.
For the final UX/UI master's work, I designed a social project in which I wanted to solve a learning problem. People don't usually have a lot of time to study. I ideated a new platform that would solve all the problems, both for Web and App.
#1 Ideation
??
People don’t have much time to study. Given the current situation of an unstable society and economy, learning according to your needs would be a great solution for most people. What if it was flexible? A platform in which we solve the two main problems of learning: Time and money.
«Profeporhoras» (Teacher by hours in spanish) is a new Learning portal in which you choose the training according to your time and budget.
- Description: Online search engine for personalized teachers by hours and topics.
- Objectives: Help people to improve their specific training with financial difficulties and who have little time.
- Target: Mainly to a target that is already working and wants to improve their profession, but does not have much time. You just need a little training. Also the second target may be more of a hobby, people who already know how to play the guitar but have forgotten it a bit, so 4 hours and everything a day.
- Needs covered: Micro-learning encourages savings; On the one hand, time and financial savings, curiously, it has always been said that time is money, so it covers the same need but in two main focuses.
#2 Architecture
The platform was designed for Web/App with a User Experience layer focused on making the user the easiest to use, with high usability and affordance. It is worth highlighting the header of the home page, which already shows an integrated search engine to make it the easiest and fastest for the user. More than 25 wireframes were designed.
- Sitemap
- Flow navigation
- Wireframes
- Design Interaction
- UI definition
#3 Web version
UI Design
I designed a complete guidelines, assets & prototype for Developers.
- Guidelines
- Brand
- Colour Palette
- Interaction
- Typography
- Headers styles (H)
- Buttons
- Iconography
#4 Mobile version
Mobile layout optimized an 1 column with left alignment.
Assets mobile & Hamburger menu included.
#5 Development
I prepared a document with all the constructive elements, system design, and assets for the developers.
#6 Prototype
Model
Need more?
Download the keynote of the Project
Saldedudas
? Oops! Sorry
The mobile version is not created.
I'm working.
Please visit on desktop version.
"Sal de dudas" a Real Estate App that find the best option to buy according your economy with machine learning.
In 2018 I studied a Master of UX/UI at EsDesign Barcelona.
"Sal de dudas" was a 2-month project of Master on I developed all the phases of a new digital product, from the ideation, generation of ideas, testing and validation of the ideas to the entire UX/UI design of the MVP.
#1 Research
??
The first step, was to work on the first subject of the project:
Digital Media Strategy. My goal was to focus getting a general idea of the current social, technological and creative situation to develop a creative process that would allow me to select and validate ideas.
- Generation of ideas
- Previous context
- Discussion and focus
- Bases and criteria
- Lateral thinking
- Exploring ideas
- Idea validation
- User interviews
- Definition
- MVP
- Business model
- Content design
#2 Definition
The idea of which was defined based on conclusive studies and user tests, was to create a social project to help people decide to know which is the best option when buying or renting their home.
An online platform that worked thanks to big data. The user had to fill out a form according to their economy and data showed them the best option. That is why the name means «Saldedudas» No doubts in spanish.
#3 Look&Feel
UI Design
I designed a complete guidelines, assets & prototype for Developers.
- Guidelines
- Brand
- Colour Palette
- Interaction
- Typography
- Headers styles (H)
- Buttons
- Iconography
#4 Architecture
I designed architecture, hierarchy & usability for the App version, focused on design user centered. In the wireframes, I added visual information about the interface, for example the elements status and all the technical indications of the design interaction.
- Sitemap
- Flow navigation
- Wireframes
- Design Interaction
- UI definition
Model
#5 Prototype
Need more?
Download the keynote of the Project
?
Rafel Gil | Senior Product Architect & Growth Strategist
Defining digital solutions
with systemic rigor
+15 years architecting high-impact products via engineering logic & curated design.
Senior Architect in Product · Design Ops · UX/CRO · Growth
Volkswagen website
Strategic UX & Usability engineering to enhance user-centricity and…
Movertis
Leading the 360° product strategy and growth architecture for a GPS fleet…
Studio Projects
8 years of cross-functional leadership for agencies and startups,…
Built for desktop.
This is a technical portfolio, system architecture, AI workflows, and a DevLog. It doesn't do it justice on mobile. Open it on your computer or send the link to yourself.
Kerning LABS®
See the latest work
These projects are part of Kerning LABS® — my product studio where business logic meets conversion design.
Every case study documents the full process: from strategy and architecture to deployment and real commercial results.
+15 YEARS OF PRODUCT & RIGOR
Human and Creative
I am a Senior Product Architect & Growth Strategist with a unique background: 12 years in civil engineering and 15 leading high-scale digital products. This trajectory allows me to merge engineering precision with strategic design to architect robust, scalable, and high-performing digital products.
Specializing in Systemic Rigor and my UX Pyramid Method®, I go beyond aesthetics to focus on product architecture, Design Ops, and UX/CRO. I don't just manage projects; I engineer products that align business goals with user needs through Applied AI and technical rigor.
I lead strategic consulting, helping companies transform complex visions into roadmaps and profitable digital products. Integrating Design Ops with systemic rigor, I ensure assets are built to scale, minimizing technical debt while maximizing ROI.
Tagline
Methodology
Change the color to match your brand or vision, add your logo, choose the perfect layout, modify menu settings, add animations, add shape dividers, increase engagement with call to action and more. Change the color to match your brand or vision, add your logo, choose the perfect layout, modify menu settings, add animations, add shape dividers, increase engagement with call to action and more.
+15
Years
+120
Projects
+2000
Pages of Brief
+10K
Cups of coffee ☕
Feedback of Rafel
15+ years merging engineering logic with curated design and UX/CRO. An approach driven by systemic rigor that consistently leaves clients surprised by its depth.
Account manager | DDB Barcelona
Rafel nos ayudó mucho en un proyecto largo y ambicioso para nuestro cliente Volkswagen. Su actitud y predisposición siempre fueron bien acogidas y valoradas.
Technological advisor at Movertis
Rafel is creative, dynamic, he is excellence. Responsible, resilient and decisive, contagious curiosity and the desire to always take a step forward, a wonderful...
UX Designer | DDB Barcelona
He’s an absolute team-player, always ready to solve problems or challenges thrown at him with sound IxD and UI proposals, thus improving the overall user...
Full picture?
Download my CV
