Rodrigo – Python, LangChain, LLM

A large-scale consumer personalization platform serving real-time recommendations and AI-powered profile features. The system processes 2+ TB of daily event data and drives user engagement through LLM-based agentic capabilities.

• Built real-time recommendation and profile-materialization services using Dataflow, Kafka/Pub/Sub, and BigQuery to serve personalized content at scale.
• Shipped 5 LLM-powered agentic features — profile coaching, image review/ranking, and bio recommendations — using LangGraph and OpenAI APIs via a RAG pipeline.
• Modernized data-quality and compliance workflows processing ~2 TB/day across 40+ event types; implemented 100+ validation rules, reducing parsing and storage costs by ~$2,000/month.
• Improved ingestion reliability through idempotent updates, conflict resolution, and end-to-end observability.

A generative AI marketing platform for a major airline that replaced agency-dependent content workflows with automated, LLM-powered personalized campaign generation. Powered 80+ ML models in production on GCP.

• Led development of an airline GenAI personalization product that replaced agency-dependent content with automated LLM-generated campaigns.
• Led migration of production infrastructure supporting 80+ ML models to Vertex AI, standardizing deployments with Kubernetes, Kubeflow, Terraform, and CI/CD pipelines.
• Managed and mentored 5 ML engineers; partnered with data, product, and business teams to productionize new features.

A personalized video news platform delivering algorithmically ranked content via hybrid recommendation models and semantic search.

• Built and deployed hybrid recommendation systems using collaborative filtering, matrix factorization, XGBoost, and LightGBM.
• Increased watch time for the top-3 recommendations by 20%, daily active users by 7%, and average session length.
• Developed feature pipelines in Snowflake and integrated Elasticsearch for semantic search and content discovery.

A real-time financial data platform built to ingest, process, and distribute banking transaction data at scale using streaming architectures on Azure.

• Built data pipelines to migrate and transform data from multiple sources into Azure Databricks.
• Built Kafka and Spark Structured Streaming pipelines into Databricks, reducing transaction-data freshness latency.
• Reduced storage and processing costs by 20% through incremental processing.

A proptech platform centralizing operational and business data for property management, analytics, and ML-driven insights on AWS.

• Built and managed a data warehouse on AWS Redshift, centralizing data from multiple operational sources.
• Deployed and managed Apache Airflow on AWS EC2 with Docker Compose as the primary pipeline orchestration layer.
• Developed ETL pipelines in Airflow to transfer and transform data from multiple sources.
• Designed and implemented ML model training and deployment pipelines to AWS S3 for production use.
• Collaborated with Data Scientists and BI Engineers to build data solutions aligned with business requirements.