I build Applied AI products and backend systems that transform real-world needs and data into functional, explainable and useful solutions.

My work combines Python, Django, Generative AI, RAG, LLMs, Machine Learning and data, with a strong focus on turning AI capabilities into working products.

I work across the development lifecycle — from understanding the problem and designing the solution to implementation, integration, testing, validation and debugging.

My previous experience in business management, product, clients and decision-making gives me an additional perspective: I understand technology not only as an engineering challenge, but as a tool that must create real value.

Applied AI · Generative AI · RAG · LLM Systems · Machine Learning · Python · Django · Backend · AI Product

An Applied AI platform for intelligent nutritional planning that combines specialized knowledge, retrieval, controlled LLM generation and deterministic validation.

Rather than relying on a single prompt, NutriPrompt separates retrieval, inference, validation, domain logic and presentation.

• Retrieval-Augmented Generation (RAG)
• Prompt Engineering
• Domain-specific knowledge retrieval
• OCR processing
• Structured JSON generation
• Business-rule validation
• Multi-provider LLM orchestration
• Gemini as primary provider
• OpenAI fallback
• Corrective retry mechanisms
• Provider fallback strategies
• Automated testing
• Debugging and root-cause analysis
• Observability capabilities

Python · Django · RAG · LLMs · Gemini · OpenAI · OCR · REST APIs · JSON

Applied AI beyond prompting: retrieval → generation → validation → actionable output.

A digital-health product exploring how longitudinal data can be transformed into contextual and useful information.

ParkinGuide is its first use case, focused on longitudinal functional monitoring in Parkinson's.

The platform is designed as a non-diagnostic system and does not provide treatment recommendations.

• Longitudinal data processing
• Baseline-relative modelling
• Contextual explainable intelligence
• Functional trend analysis
• Data preparation and exploration
• Machine Learning modelling
• Data visualization
• Privacy-oriented architecture

Machine Learning techniques are selected according to the characteristics of the data and the problem.

ParkinGuide currently includes a Random Forest Regressor as one concrete implementation within the wider intelligence architecture.

MAE: 3.59 · RMSE: 4.84 · R²: 0.946

Python · Django · Pandas · NumPy · scikit-learn · Machine Learning · SQLite · Chart.js

From longitudinal data to contextual intelligence.

A full-stack real-time platform developed end-to-end, covering backend architecture, relational data modelling, authentication, matching logic and real-time communication.

• Flask MVC backend
• User registration and authentication
• JWT-based login
• Password recovery with one-use tokens
• bcrypt password hashing
• User CRUD
• Image upload and validation
• Interests and matching logic
• Likes and matches
• Geolocation and distance
• Private real-time chat
• Persistent message history
• Typing indicators
• Real-time notifications
• WebSocket communication
• PostgreSQL relational data model
• Database indexes
• Dockerized architecture
• NGINX reverse proxy
• Persistent volumes
• Environment-based configuration
• Health checks and restart policies

Python · Flask · PostgreSQL · Flask-SocketIO · Docker · NGINX · JavaScript · HTML · CSS

A complete backend and real-time product built from architecture to deployment.

Generative AI · LLMs · RAG · Prompt Engineering · Machine Learning · OCR

OpenAI · Gemini · scikit-learn

Pandas · NumPy · PostgreSQL · SQLite · JSON

Python · Django · Flask · REST APIs · Flask-SocketIO

Git · GitHub · Docker · NGINX

HTML · CSS · JavaScript

Automated Testing · Validation · Debugging · Root-Cause Analysis · Corrective Retry · Fallback Strategies

42 has been a key part of my journey into software development.

Its project-based, peer-to-peer model strengthened the way I approach technology today:

build → test → fail → debug → understand → improve

Beyond programming, 42 taught me to work through complex problems autonomously, collaborate with peers, review other people's work and keep learning when there is no predefined solution.

Problem Solving · Autonomous Learning · Peer Collaboration · Ownership · Continuous Iteration

I remain actively involved in the 42 Urduliz community, combining technology, mentoring and ecosystem building.

My current contribution includes:

• Technical mentoring in Python
• Supporting practical and collaborative learning
• Problem-solving support
• Connecting talent, companies and the technology community
• Networking and communication across the ecosystem

For me, 42 is not only where I strengthened my technical foundations.

It remains part of how I learn, collaborate and contribute to the technology community.

Artificial Intelligence evolves quickly, so continuous research and experimentation are part of the way I work.

I explore emerging architectures and evaluate them against a simple question:

Does this improve the product or solve the problem better?

Current areas of exploration:

Agentic AI · AI Agents · RAG · Multi-Agent Systems · Applied AI

I recently completed IBM SkillsBuild — Make Agentic AI Work for You, expanding my understanding of AI agents, RAG and multi-agent approaches.

My path into AI is not purely technical.

Before moving into software development, I spent many years working across business management, commercial strategy, finance, clients, teams and product.

Today, I combine that experience with software and AI development.

It means I naturally approach technology through three connected questions:

For me, successful AI is not simply about choosing the most powerful model.

It is about designing the right system around the problem.

Technology should create clarity, not noise.

I believe in:

Human-centered technology · Applied AI with purpose · Building before theorizing · Continuous learning · Real-world impact

Applied AI · Python · RAG · LLM Systems · Machine Learning · AI Product

Building AI systems for real-world problems.