Diego Pino

Liderazgo técnico de varios proyectos de IA, destacando el chatbot de ia generativa de la página principal de iberia

I act as technical lead for Generative AI initiatives at Iberia, focusing on the execution and evolution of the main customer-facing conversational platform deployed on Iberia’s public website.

The solution is a production-grade, multi-agent Generative AI system, handling 3M+ customer conversations per year, and integrated with core operational systems through MCPs.

Key responsibilities and contributions (Generative AI):

-Technical leadership of implementation teams, translating business decisions into production-ready solutions

-Active contribution to the design of the multi-agent GenAI architecture, including agent orchestration, shared memory, and API contracts

-Hands-on development of core agents (security, tone, translation, conversation management, flight cancellation)

-Design and maintenance of a shared Generative AI core library, reused across multiple initiatives

-LLMOps execution, including deployment pipelines, monitoring, evaluation, and observability

-Close collaboration with internal Iberia technical, product, and business stakeholders, contributing to technical decision-making and roadmap discussions

Additional responsibilities (Traditional ML & Customer Analytics):

-Technical supervision of 3 Data Scientists working on production ML models, including:

-Next Best Action

-Business vs Leisure segmentation

-Churn and purchase propensity models

-Recommendation systems

-Architectural and quality oversight of end-to-end ML pipelines, from data ingestion to deployment and monitoring

-Support in incident resolution, performance analysis, and continuous model improvement

Tech stack: Python, LLMs, Agent-based architectures, AWS (SageMaker, Fargate, DynamoDB, RDS, S3, Redis), Airflow, REST APIs, OpenTelemetry, Langfuse, CI/CD

I created and put into production end-to-end churn, propensity and recommendation models. I have designed a recommendation bot by combiningfoundational models (GPT-3.5, Llama2, Claude), LangChain architecture, and prompt engineering, which interacts with a database to report available flight paths. ETL, SageMaker pipelines, CI/CD development AWS SageMaker, Apache Airflow, Model deployment, AWS, PySpark, Airflow