• What I do: Co-Founder and Technical Lead at MART Automations, where we build AI automations for companies in Panama. RAG systems, agents, and pipelines that have to survive real data.
• What I'm studying: Systems Engineering.
• What I'm into: the unglamorous half of AI engineering. Evals, failure modes, cost per run, and what happens on day two.
• Languages: Spanish (native) · English (B2) · French (B1)

I teach the engineering ideas behind useful AI systems, then give people a small exercise they can run and adapt for their own study. Each talk below has its own public resource repository with a Colab notebook, PDF summary, workshop brief, and usage notice.

The FISC workshop goes from tools, MCP, skills, and memory to a working Pydantic AI agent. InnovaTech connects context and grounding to an AI workspace. The UTP talk for the Peruvian delegation shows how an academic workflow grows from a chatbot into an assistant. EurusConf takes a LangChain agent from Colab toward a local project. The earlier UTP workshop covers prompt design and the first agent loop.

The public repositories contain participant resources. My slides and speaker notes are separate. Each repository explains its usage terms; permissions already granted for older CC BY 4.0 versions remain in effect for those versions.

• FinWise — FastAPI + Gradio for querying financial PDFs in natural language. Full RAG pipeline with semantic chunking, FAISS indexing and source-referenced answers.
• rtovar-aios — a template for students who want to run their own AI operating system. MIT.
• Inventory agent for a private client (MART Automations) — worked on a pilot that lets a team report stock movements by text or voice in Telegram, consult a shared inventory, and request summaries. It connects conversational input to a structured record with checks before changes are accepted. The client system and source repository are private.