© 2026 Iván Caamaño.

Lead automation initiatives that blend AI agents, RAG systems, and computer vision for consulting engagements.

• Designed and deployed AI-powered document-generation agent systems using Amazon Bedrock, Lambda functions and Claude Sonnet, reducing production time from 8h to 24min per document (95% reduction).
• Engineered RAG implementations for the development of chatbots using Azure Cognitive Search, ML and AI Foundry, using vectors and embeddings to decrease information retrieval latency by 90%.
• Developed AI-based solutions for the segmentation and analysis of images using GCP and Vertex AI, designing the end-to-end pipeline and cutting workload in classification by 80%.

Supported BBVA's Risk team delivering the Internal Model Method (IMM) platform with production-ready tooling.

• Designed and implemented ETL pipelines for 4 IMM risk calculation processes, processing 2M+ credit exposure records daily using Python (Pandas, PySpark) and Java batch jobs on Unix environments.
• Deployed microservices architecture using Docker and Kubernetes for parallel processing of EAD calculations, reducing batch execution time by 10%.

Coordinated with a team of 19 students for the realization of the Employment and Technology Forum of the University, with +100 participants and +20 collaborating companies.

• Coordinated +100 participants each year, including companies like Apple, Amazon and BBVA, and organized +80 conferences, workshops, and activities arranged in the months leading up to the forum.
• Developed features and improved the UI/UX of Satelec's website, increasing general web interaction by 60% and CV submissions by 40%.

Contributed ML and data-processing capabilities to modernize AESA radar programs tied to Eurofighter platforms.

• Developed CNN-based model for radar signal classification using PyTorch, training on 50K+ simulated AESA radar returns to detect and classify aerial targets improving pattern detection efficiency by 30%.
• Implemented noise reduction pipeline using TensorFlow, applying wavelet transforms and autoencoder architecture to filter Doppler radar data, enhancing signal-to-noise ratio from 12dB to 18dB.