AI Engineer for production delivery
Production-ready AI systems for teams moving beyond prototypes¶
I help teams turn AI pilots into reliable systems using RAG, APIs, and agent-based workflows.
4+ years across data science, data engineering, and AI platforms, with hands-on delivery in agentic AI, retrieval, and production APIs.
Problems I help solve¶
Common gaps when an AI demo needs to become a dependable system.
-
The demo works, but production still feels fragile
The model looks good in a notebook, but the workflow breaks once routing, retries, human review, and system boundaries are added.
-
Retrieval quality is inconsistent
Vector search misses exact business terms, keyword search misses intent, and answers drift too much to trust.
-
No one is shaping the production path
Product, data, and engineering all see the opportunity, but nobody is defining the architecture that turns it into a reliable service.
What I do¶
I focus on making AI systems reliable, measurable, and easier to operate.
-
Agentic AI systems
Deterministic workflows with LangChain, LangGraph, PydanticAI, Agno, and MCP so multi-step behavior stays observable and testable.
-
RAG and hybrid search
Retrieval systems that combine vector and keyword search, improve domain accuracy, and keep the LLM layer loosely coupled.
-
Production AI APIs
FastAPI services, async processing, streaming interfaces, and deployment patterns that support real usage instead of demo traffic.
Featured case studies¶
Three recent projects with measurable production outcomes.
-
AI-Powered Email Automation System
Production GenAI workflow that reduced daily email triage from 100+ items to 10-15 actionable items using PydanticAI, RAG, and FastAPI.
-
AI-Powered Booking Email Automation
Orchestrated LLM workflow on AWS that automated booking email processing across 4 categories with idempotent supplier communication and self-service operations.
-
Conversational AI Seating Agent
Real-time conversational backend that processes Excel booking files via WebSocket, executes deterministic seating assignments, and answers operational queries in natural language.
Need help turning a useful AI prototype into a production system?¶
If the business case is already clear but the delivery path still feels fragile, I can help define the architecture, the delivery risks, and the next implementation steps.
Watch the technical talk¶
AI-Powered Email Automation: From Chaos to Action
A public walkthrough of a production email automation system that reduced daily triage from 100+ messages to 10-15 actionable items.
What an engagement usually looks like¶
-
1. Clarify the target system
We align on the business outcome, where AI belongs in the workflow, and how success will be measured.
-
2. Build the right architecture
I design the workflow, retrieval layer, contracts, observability, and deployment model around real operating constraints.
-
3. Harden for handoff
The outcome is a system your team can run, extend, and evaluate with confidence.
Certifications and speaking¶
-
Microsoft Certified: Azure AI Engineer Associate
Azure AI solution design and delivery across language, knowledge, and automation workloads.
-
Microsoft Certified: Azure Data Scientist Associate
Azure machine learning experimentation, training, deployment, and model operations.
-
Databricks Certified Machine Learning Associate
Machine learning delivery using Databricks and the Apache Spark ecosystem.
-
Technical speaker - Datamecum Webinar 2025
Public session on a production email automation system built with PydanticAI, RAG, and FastAPI.
Frequently asked questions¶
What kind of projects are you best suited for?
Projects where AI already looks useful, but the team still needs a production-ready architecture. That includes agentic workflows, hybrid RAG, and AI APIs serving real users.
Can you work with an internal engineering team?
Yes. I can work independently or as a technical partner embedded with an internal product or engineering team.
How do you approach delivery?
I start with the business goal and success metrics, then design around reliability, observability, and maintainability.
Where are you available?
I am based in Spain and available remotely across Europe. I work in English and Spanish.
How does pricing work?
Most work is project-based with clear milestones. Ongoing support can be structured as a retainer after the initial call.
Ready to evaluate your AI backlog?¶
Bring the current problem, the stack you already have, and the delivery constraints that matter. I will help map what should be automated, what should stay deterministic, and what it will take to ship it safely.