Adrián Pabón Mendoza

About Me

👋 Hey! I am Adrián Pabón, an electronic engineer from Barranquilla, Colombia, with 3 years of experience in applied AI. I specialize in language models, computer vision, and cloud-based products for real users.

I work across the full lifecycle, from idea and prototyping to deployment and iteration, collaborating with multidisciplinary teams to deliver robust and scalable solutions.

I have worked across research and industry, from scaling AI assistants for 13,000+ users to deploying computer vision pipelines that reduced manual camera review by 80%.

Experience

• Senior Data and AI Engineer · nyxn

  Leading the migration of core banking procedures to GenAI-optimized microservices and event-driven architectures. Designing serverless AWS solutions (Lambda, Step Functions, Bedrock) for secure, scalable AI workflows.

  Remote, Colombia · Feb 2026 - Present

• AI Product Specialist / AI Engineer · Optimize IT SAS

  Improved theft detection with computer vision, reducing manual review by 80%. Delivered LLM/RAG APIs with embeddings + vector search and automated OCI backup workflows with serverless functions.

  Barranquilla, Colombia · Jul 2024 - Feb 2026

• AI Researcher & Developer · Universidad del Norte

  Built and scaled NAIA to 13,000+ users. Integrated OpenAI LLMs, Redis state management, Azure services, and Microsoft Graph APIs to provide secure assistant capabilities in 3-5 seconds.

  Barranquilla, Colombia · Jan 2024 - Nov 2025

Projects

‹ ›

NAIA

Multi-role assistant for Universidad del Norte that combines LLMs, RAG, Redis state, Azure services, and Microsoft Graph integrations for academic and administrative workflows.

2024-2025 · 13,000+ users · LLMs · RAG · Redis · Azure · Link to the project · Watch Demo

• AI Agent Implementation for Medical Call Center

  Intelligent voice agents for outbound campaigns in a clinical laboratory. Implemented tool-calling logic for real-time appointment and lab result validation connected directly to medical databases.

  2026 · Impact: 50% reduction in routine call handling time

• Theft Detection (Computer Vision)

  Production CV pipeline for micro-market security, including dataset curation, labeling, model training/validation, and hold-out evaluation.

  2024 · Impact: 80% reduction in manual camera review time

Research

• NAIA: A Robust AI Framework for Academic Assistance

  Authors: Adrián Pabón Mendoza, Kenneth Barrios Quiroga, Samuel Solano Celis, Christian G. Quintero M.

  Systems (2025) · View paper

• A Comparison of Different Architectures for AI-Powered Virtual Assistants

  Authors: Kenneth Barrios Quiroga, Adrián Pabón Mendoza, Samuel Solano Celis, Christian G. Quintero M., César Viloria-Nuñez

  IEEE TEMSCON LATAM (2025) · View paper

• NAIA: A Multi-Technology Virtual Assistant

  Authors: Kenneth Barrios Quiroga, Samuel Solano Celis, Adrián Pabón Mendoza, Christian G. Quintero M., César Viloria-Nuñez

  IEEE Access (2025) · View paper

• Looking for operational efficiency in itinerary assignment optimization: a comparison of different optimization approaches

  Authors: Adrián Pabón Mendoza, José Oñate López, Christian G. Quintero M.

  IEEE COLCOM (2024) · View paper

Engineering Notes

Short ideas I would expand into posts: practical lessons from building AI systems for real users.

• AI assistants

  Lessons from scaling an AI assistant to 13,000+ users

• RAG systems

  Why retrieval quality matters more than prompt tricks

• Computer vision

  Computer vision projects fail before model training

Interests & Skills

Contact