AI Engineer · Co-founder · Technical lead
I work with software teams to define, build and integrate AI systems. I take responsibility for architecture and the technical criteria that guide delivery, with a focus on private AI, evaluation and product development.
Bilbao / Bizkaia · hello@tenbeltz.com
My work combines software engineering with designing and evaluating AI systems, including environments with sensitive information. At TenBeltz I lead consulting and technical definition, and coordinate the team’s development and integration work.
Since June 2025
Software development and technical contributions to AI projects. Architecture, model selection and integration for Konect call analysis, including a local AI adaptation for RETA on NVIDIA DGX Spark.
Since July 2025
Architecture and development of a child protection platform. Backend, private infrastructure, local AI and access controls. Led and implemented ENS Medium compliance.
Since August 2024
Consulting, technical definition and delivery criteria for SaaS teams and software consultancies. Voice agents, document processing, report generation and product integration, coordinating the team’s development work.
June 2024 — June 2025
Interfaces and software components for SISLOG warehouse management. Spring and TypeScript development, Python REST integration with AI services and a Java connectivity library.
February 2023 — August 2024
Python and Django applications, process automation, web scraping and bots for international clients.
Transcription, LLM analysis and quality metrics, with a local AI adaptation for RETA.
Voice agents connected to the restaurant’s CRM, tables and opening hours.
From all-LLM classification to semantic features and a lightweight supervised ensemble, for hundreds of thousands of documents per month.
Case files, documents and reports in a platform with local AI, access controls and traceability.
Structured extraction, drafts and quality checks over clinical documentation.
I account for integration, errors and maintenance from the design stage. Delivery includes the information your team needs to operate the system.
I define representative cases and acceptance criteria to assess results and detect regressions.
Logs, traces and monitoring help explain what happened and support diagnosis when something fails.
I evaluate model choices, architecture and cost per task alongside quality and performance.
30 hours of local AI training delivered through Imagina to Opus Dei’s international technical team. Model serving with vLLM, private architectures, evaluation, RAG and agents.
First prize in BIC Bizkaia’s business ideas competition with SustainaAI, a project combining sustainability and artificial intelligence.
CIFP Txurdinaga LHII · 2023—2025
Multiplatform Application Development (DAM).
42 Urduliz · 2024—2025
Project-based programming, C and collaborative learning.
Spanish: native · Basque: advanced · English: professional working proficiency.
RAG · LangGraph · Typesense · Qdrant · vLLM · Ollama
Python · FastAPI · Django · TypeScript · React / Next.js · PostgreSQL
Docker · Redis / ARQ · Langfuse · LLM-as-judge · Grounding
AI diagnosis · 2 min
TenBeltz