Tito Lasanta Viñes

Designed and developed 4 AI systems, including a dashboard-generation system (85% success rate) and an integrated expert-consulting system (92% accuracy, 500+ consultations/week), 2 sold to external clients. Built RAG pipelines and vector search infra (pgvector, Qdrant). Built event-driven pipelines with Kafka and Spark; used Databricks in production. Managed PostgreSQL, MongoDB, Redis; Airflow/Prefect orchestration. Docker/Kubernetes with CI/CD.

Led design, development, and deployment of Zonia.com end-to-end, including architecture decisions for scalability. Built responsive front-end with React.js and TypeScript; back-end with Node.js integrating SQL databases. Optimized concurrent response handling for data retrieval and recommendations. PostgreSQL data layer, Redis caching, Docker containerization, CI/CD.

Owned scoutsmartrecruit.com and its companion mobile apps end-to-end: built the back end (Entity Framework, C#/.NET) and a responsive React.js front end. Owned the recommendation system combining hardcoded rules with a CNN-based model, backed by PostgreSQL and MongoDB; basic Airflow scheduling, analytics on Snowflake. Handled client communication and data cleansing; CI/CD and Docker-based deployment.

Worked on a research project involving AI applications and code translation for climate data analysis. Processed large-scale climate datasets with Apache Spark; analytical queries on BigQuery; pipeline scheduling with Apache Airflow; exploration and modeling in Databricks.

Thesis: Transformer model with divided attention and recurrent layers applied to predicting the El Niño phenomenon. Designed and trained in PyTorch: divided attention separates spatial and temporal attention computation, recurrent layers capture longer-range temporal dynamics. Achieved 0.72 Pearson correlation and 0.6 RMSE at a 1-year horizon, across 10M+ multidimensional data points.